It has been known for many years that administration of corticosteroids can cause atrophy of the adrenal glands (Ingle and Kendall, 1937), and that the severity of the atrophy is proportional to the duration of administration (Salassa, Bennett, Keating, and Sprague, 1953; Bennett, 1954). Kilby, Bennett, and Sprague (1957) described degranulation of the basophil cells of the pituitary which also correlated in severity with the duration and the amount of corticosteroid administered. That these structural changes may be accompanied by severe functional inadequacy became apparent after the publication of reports of irreversible postoperative shock developing in patients who had previously received corticosteroids (Fraser, Preuss, and Bigford, 1952; Salassa and others, 1953; Bayliss, 1958; Winstone and Brooke, 1961). Salassa and others pointed out that only a small minority of patients at risk developed shock of this kind, and that it was impossible to foresee this complication in any individual patient. Winstone and Brooke regretted "the absence of a completely dependable test of adrenal reserve" which would enable those patients with pituitary-adrenal insufficiency to be detected. The importance of reliable objective tests of pituitary-adrenal function in patients who are being considered for withdrawal of steroid therapy, or from whom steroid therapy has been recently withdrawn, is thus apparent. Treadwell, Savage, Sever, and Copeman (1963) showed that, in a series of 41 patients who had received corticosteroids, less than half showed an adequate response to metyrapone, yet all showed a good response to ACTH. The metyrapone test, however, may not be very suitable in this context for it only indicates the response of the pituitary to the stimulus of a low plasma cortisol by a "feed-back" mechanism (Liddle, Island, Lance, and Harris, 1958). Failure to respond to metyrapone might not mean inability to respond to stress. There is evidence that the "feed-back" mechanism differs from the stress response mechanism, and James and Landon (1964) have shown that, when the plasma cortisol is reduced to zero by means of an infusion of dexamethasone, it will still rise promptly in response to the stress of hypoglycaemia. Amatruda, Hollingsworth, D'Esopo, Upton, and Bondy (1960) and Livanou, Ferriman, and James (1965) have used insulin-hypoglycaemia to assess the ability of the pituitary to respond to stress following corticosteroid therapy. This may be more useful than the metyrapone test in predicting the development of any serious pituitary-adrenal insufficiency. This paper reports the results of pituitary-adrenal function tests performed during an attempt to withdraw corticosteroid therapy from a group of patients with rheumatoid arthritis.