The potential effects of growth hormone (GH) deficiency in adults and the importance of GH secretion in adult life have only been recognized and documented recently. It has been suggested that GH-deficient adults may have premature mortality, abnormalities in body composition and bone density with impaired physical performance and psychological well-being, which are sometimes improved by GH replacement. It is essential, therefore, to establish reliable standards to define GH deficiency in adults. Patients with possible GH deficiency often have primary pituitary or hypothalamic disorders or have undergone surgery or radiotherapy, and thus show evidence of a failure of one of the other pituitary hormones. Several biochemical approaches have been studied to define GH deficiency in the adult and no universal consensus has yet been reached. The most widely established criterion is the peak serum GH concentration achieved during a provocative test, usually the insulin tolerance test (ITT), or following other pharmacological stimuli (e.g. glucagon, arginine, clonidine or GH-releasing factor) but, alternatively, a more physiological stimulus (such as sleep, fasting or exercise) has been used. Spontaneous circulating levels of hormones of the GH axis [24-hour integrated GH concentration, serum insulin-like growth factor I (IGF-I) or IGF-binding protein-3] have been used in the diagnosis of childhood GH deficiency. They have been tested in adults as well but seem to have a more limited role. There are several factors complicating the evaluation of these results. Basal and stimulated GH and IGF-I levels decline with age and with obesity, levels tend to be higher in females and are dependent on nutritional and physical status. The ITT potentially has some risk attached, e.g. in the presence of ischaemic heart disease, but it has proved to be safe in general when used in specialized departments. Other tests are less reliable; releasing hormone tests only assess the readily releasable stores within the pituitary and not the physiological secretory status. The 'cut-off' point for the definition of subnormal responses ideally needs to be set for each provocative test, for each age group, for each degree of obesity and for both sexes. There is considerable variability in GH assays among different laboratories, which makes it difficult to compare hormone levels. The reproducibility of provocative tests can also be variable. An advantage of the hypoglycaemia and glucagon tests is that they allow simultaneous assessment of the adrenocorticotropic hormone reserve.