The reproducibility of measurements of nocturnal spontaneous secretion of growth hormone (GH) was assessed. The study population consisted of 15 short children with a variety of pathological conditions. Blood samples were taken every 20 min from each subject during sleep on two consecutive nights. The analysis of variance of matched series indicated a global reproducibility, but also demonstrated significant individual variabilities (within-subject effect) of both peak amplitude (p = 0.05) and mean concentration (p = 0.02). We did not find any link between the variation of the GH parameters and the variations of the clinical sleep score in 11 patients.