Possible causes of error in the diagnosis of isolated GH deficiency are the variability of GH response to repeated tests, the existence of transient GH deficiencies, and the low GH levels found in short statured children with delayed puberty. Sixty-three patients with variously expressed GH deficiency were retested (1 sleep test and 2 pharmacological tests) after 1-3.9 yr of GH therapy (dose, 15 U/m2.week). Forty-eight subjects had arginine, L-dopa, and sleep tests (mean serum GH concentration) twice, while 15 had only arginine and L-dopa tests. All patients were retested 1 month after withdrawal from therapy. The criteria used to subdivide the patients were pubertal development and response to pharmacological and sleep tests at first diagnosis and on retesting. The initial diagnosis in 33 subjects (52.4%) was not confirmed, and 13 (20.6%) were no longer deficient on retesting. The percentage of normalization was high for the sleep test (43.9%), lower for the pharmacological test (24.5%), and lower still (12.9%) for pharmacological and sleep tests considered together. While none of the 28 subjects who remained prepubertal at retesting normalized in any of the tests, 13 of the 35 subjects retested during puberty did. When normalization was observed in pubertal subjects, it occurred predominantly in the sleep test. Growth velocity and height age/bone age increment ratio after the first year of therapy were no different for the groups of subjects classified according to GH secretion on retesting. Our study demonstrates that a number of children diagnosed as GH deficient do not have a true deficiency. However, such a diagnostic error seems to have little effect, at least in the first year of therapy, on the effectiveness of GH treatment.