Radical transsphenoidal surgery in 58 patients with large non-functioning pituitary tumours relieved pressure symptoms and disconnection hyperprolactinaemia without loss of pituitary function, and verified the diagnosis. Pre-operative bromocriptine did not cause any shrinkage of non-functioning tumours (n = 8), even up to 48 weeks. Thirty-two patients not given postoperative radiotherapy are being followed-up, and although two show minor CT scan evidence of tumour regrowth, for the others the delay in radiotherapy is enabling reproduction where required. In 20 patients with non-invasive macroprolactinomas radical surgery caused no loss of pituitary function but cured eight, of whom one has relapsed; none of eight invasive prolactinomas was cured by surgery. Pre-operative bromocriptine caused marked size reduction in seven of seven macroprolactinomas, but if continued beyond 6 weeks induced tumour fibrosis and uneven shrinkage which made surgery dangerous and unproductive. Selective transsphenoidal surgery relieved hyperprolactinaemia in 70% of patients with meso- (n = 15) or microprolactinoma (n = 24) usually without loss of pituitary function: the relapse rate was 1 per 88 patient years of follow-up. Invasive prolactinomas, and macroprolactinomas showing uneven shrinkage with a short (up to 4 weeks) course of bromocriptine should have radiotherapy rather than surgery, while the patients most suited to surgery are those with mesoprolactinomas, or some macroprolactinomas with compact intrasellar shrinkage after short-term bromocriptine. Microprolactinomas generally did not benefit from surgery as compared to conservative therapy. In none of the tumour groups were the surgical findings an accurate predictor of postoperative pituitary function.