Observer variation in interpreting sellar radiographs in patients suspected or known to have a pituitary tumor has been examined. Two radiologists experienced in interpreting sellar radiographs examined independently, without clinical details, plain films and tomograms of the sella of 101 patients. In most, only minor changes were anticipated. Of the 93 female patients, 67 were under investigation for amenorrhea. Radiographs were examined four times, each radiologist examining each set twice. Appearances were classified as normal, doubtful or abnormal on each occasion. Overall intraobserver agreement was 76%--85%. Neither radiologist changes his opinion by more than one category, e.g. from normal to doubtful. Overall interobserver agreement was 63%--75%. Disagreement between observers concerning 11 (11%) of the patients resulted from differences of opinion about whether minor changes in sellar outline represented an abnormality or merely a normal variation. Kappa analysis suggested that much of the agreement may be ascribed to chance. Agreement rates resemble those for other clinical and radiological investigations.