Because current concepts of growth hormone (GH) testing and GH treatment have become controversial, we investigated the GH secretory patterns in children with normal and short stature. Twenty-four-hour serum GH levels were evaluated in three groups of children. Group 1 was composed of children with normal height (mean height = 0.02 SD, n = 33); group 2 was composed of short children (less than 5th percentile, n = 63) with normal results on provocative GH testing; and group 3 was composed of short children (less than 5th percentile, n = 7) with subnormal results on provocative GH testing. Mean +/- SD (range) GH levels during 24-hour studies of GH secretion were 1.6 +/- 1.1 (0.5 to 5.6), 1.8 +/- 1.2 (0.6 to 6.3), and 0.9 +/- 0.4 (0.5 to 1.7) ng/ml in groups 1, 2, and 3, respectively. No statistical difference existed in mean GH levels between groups 1 and 2 or between groups 1 and 3. The mean GH concentration from 24-hour studies in group 2 children did not correlate with chronologic age, height standard deviation, growth rates, or insulin-like growth factor 1 levels. The linear growth rate of 26 of 28 children in group 2 who received GH therapy for 6 months improved by 2 cm/yr or more; the mean +/- SD growth rate was 4.0 +/- 1.3 and 8.8 +/- 2.0 cm/yr during control and treatment periods, respectively, for these 28 children. Mean GH levels from testing did not predict response to GH during 6 months of therapy. Children with slower growth rates responded better to GH therapy (p less than 0.05). We conclude that (1) in 24-hour studies, GH levels in normal children overlapped with those of short children, including those with classic GH deficiency, (2) in 24-hour studies, GH levels did not predict responses of linear growth to short-term GH treatment, nor did they correlate with children's heights or growth velocities, and (3) the majority of short children in group 2 treated with GH for 6 months had an increase in linear growth velocity, the mean +/- SD change being 4.8 +/- 2.0 cm/yr.