Forty-nine children with short stature (age range, 4.1-15.9 yr) were examined. Twenty-four (group 1) were submitted twice to an arginine and a sleep test (12-h overnight GH profile). Twenty-five patients (group 2) were submitted twice to an arginine and L-dopa test. Coefficients of variation were calculated between both the results of pharmacological (peak and area under the curve) and sleep tests [mean GH concentration (MGHC), peak, area under the curve, number of peaks above 5 micrograms/L, and peak area]. In group 1 the coefficient of variation of sleep test parameters was significantly lower than that of pharmacological tests (P less than 0.01 to less than 0.001). In the sleep test the area under the curve and MGHC were the most constant parameters. Group 2 showed no difference between the coefficients of variation of the two pharmacological tests. Considering groups 1 and 2 together, the coefficients of variation of the sleep test, in particular the MGHC and area under the curve, were lower than those of the two pharmacological tests. Eight of 24 subjects in group 1 showed a low GH level in 1 series of tests, and a normal level in the other series. Five of 18 subjects in group 2 showed an abnormally low GH response to the arginine and L-dopa tests and a normal response to the 2 repeated tests. Therefore, to prevent an erroneous interpretation of the GH test results, it is very important to perform a sleep test and repeat it whenever GH secretion seems to be deficient or at the lower limits of normalcy.