A total of 5,473 pharmacological provocative growth hormone release tests were carried out in 3,143 children. Mean age was 9 years 9 months (range 3-16 years) and mean bone age was 7 years 6 months (range 2-14 years). Tests were of 9 different types: 1) arginine (n = 625); 2) clonidine (n = 339); 3) insulin (n = 198); 4) ornithine (n = 162); 5) insulin + arginine (n = 203); 6) clonidine + betaxolol (n = 2,003); 7) L-dopa (n = 685); 8) glucagon = propranolol (n = 443); 9) glucagon + betaxolol (n = 815). All growth hormone determinations were performed using the same radioimmunoassay. Distribution of values obtained with each test was gausso-logarithmic. Mean peak levels with their 95% confidence limit were as follows: 1) 10.2 and 0.45; 2) 11.5 and 0.7; 3) 11.8 and 0.8; 4) 14.2 and 1.2; 5) 14.3 and 0.9; 6) 15.7 and 1.1; 7) 19.8 and 2.1; 8) 20.8 and 2.3; 9) 21.0 and 2.5. These data indicate low specificity, with up to two-fold differences in mean peak levels from one test to another; proportions of peaks under 10 ng/ml ranged from 29% to 69%. Thus, the rate of patients diagnosed with growth hormone deficiency may vary substantially according to the test used. To reduce these discrepancies, we suggest adjustment of test results using a weighting coefficient of 1) 1.9; 2) 1.48; 3) 1.4; 4) 1.16; 5) 1.06; 6) 1.01; 7) 0.73; 8) 0.69; 9) 0.66.