A phantom study was performed to assess the accuracy of MRI in volume determination of pituitary macroadenomas, using the summation of areas technique. Five phantoms, 3-4 cm in diameter, simulating pituitary macroadenomas of various shapes, were evaluated. The volume could adequately and reproducibly be determined, regardless of minor rotational variations of the phantom position. Using a three-dimensional SE sequence with 2.5 mm slices, the error was independent of the shape of the phantom and did not differ significantly from zero (less than 5%); this sequence was found to be the best of those tested. The coefficient of variation between examinations with different tilts was 2%. In a clinical part, the volume of pituitary macroadenomas was determined with the same technique on coronal two-dimensional 5 mm images and was used as a "gold standard" to which the largest transverse, sagittal, vertical, and oblique diameters, a central tumor area, and the product of the three orthogonal diameters were correlated. In interpatient comparisons the largest diameter was found to be useful, but only as a rough measure of the tumor size. The formula 0.5 x width x length x height provided a fairly adequate estimation of the tumor volume, except in the largest tumors. In intrapatient comparisons during bromocriptine treatment, the height was found to be the best of the diameters, but the central tumor area or the product of the orthogonal diameters was preferable.