Controversy surrounds the reliability of high-resolution computed tomography (CT) in the identification of functioning microadenomas, especially in patients with hyperprolactinaemia and hypercortisolaemia. We studied 34 such patients: each underwent a high-resolution CT examination of the pituitary gland and each had a trans-sphenoidal exploration. Computed tomography showed 28 focal abnormalities but only 20 of these were confirmed to be due to a tumour at operation. In the other eight the tissue at the site of the radiological abnormality was normal histologically. In a total of 29 patients a tumour was found at operation and confirmed histologically; 20 were at the site of an abnormality shown by CT but in nine a tumour was found in a part of the gland judged to be normal radiologically. In general, good correlation between CT and the surgical findings was found only with tumours more than 6 mm in size. Of the indirect signs of a tumour, such as stalk deviation and a convex upper surface of the gland, only the former showed any correlation with the presence of a tumour. The results show that cautious, even sceptical interpretation of pituitary CT scans is essential and suggest that this investigation is not sufficiently reliable to be used as a routine in the assessment of hyperprolactinaemia or Cushing's syndrome. Computed tomography should be reserved for patients undergoing pituitary surgery, those with an enlarged fossa on plain radiography and those with extreme hormone hypersecretion.