84 short children were submitted to nocturnal spontaneous growth hormone (GH) secretion tests and to provocative insulin-arginine tests. Discrepancies between the GH peak under the provocative test (I-AP) and nocturnal GH maximal peak (PA) and mean concentration (MC) were frequently observed, despite significant statistical correlation between I-AP and PA (r = 0.47; p < 0.02) and between I-AP and MC (r = 0.42; p < 0.02). Night profiles were evaluated by time analysis: 31 fitted a theoretical model, consisting of a cosine function of time (modelizable profiles). Spectral analysis, from Fourier transformation, indicated predominant periods after cluster analysis. The major predominant period in modelizable (n = 9) and in nonmodelizable (n = 28) profiles was close to 180 min and a secondary period was on average 122 min in modelizable (n = 20) and 105 min in nonmodelizable (n = 23) profiles. Two modelizable and two nonmodelizable profiles escaped this classification. The general, auxological and GH secretory status did not differ significantly between patients with modelizable and nonmodelizable profils. Growth velocity correlated with GH mean concentration (r = 0.36; p < 0.001), but not with plasma insulin-like growth factor-I levels nor with any of the pulsatility indices: number of peaks, main period, and pulse height index = mean GH peak/mean GH concentration. The relevance of GH pulsatility to growth is, therefore, unclear in humans.