Urinary growth hormone (uGH) excretion and serum growth hormone concentrations have been compared in three groups of children. Group 1 consisted of 21 children who had had cranial irradiation as part of their treatment for acute lymphoblastic leukaemia; group 2, 18 normal children; and group 3, 12 boys with constitutional delay in growth and puberty who were in early puberty. Children in groups 1 and 2 each had a 24 hour serum growth hormone profile (sampling every 20 minutes) and concurrent urine collection. The 12 boys in group 3 had a total of 21 profiles (sampling every 15 minutes for 12 hours) and concurrent urine collections. In the prepubertal children (n = 17), in both groups 1 and 2, there was a significant correlation between mean serum growth hormone and total uGHng/g creatinine. There were also significant correlations between total uGHng/g creatinine and both peak serum growth hormone and mean amplitude of the pulses in the growth hormone profile. In the pubertal children (n = 22), in groups 1 and 2, whether combined or in separate groups, there was no significant correlation between total uGHng/g creatinine and mean serum growth hormone, peak serum growth hormone, or mean amplitude of the pulses in the growth hormone profile. In group 3 there were significant correlations between total uGHng/g creatinine and both the mean serum growth hormone and mean amplitude of the pulses in the profile. Therefore uGH estimations appear to correlate well with serum growth hormone profiles in children who are prepubertal or in early puberty, but not in those further advanced in pubertal development. These results may reflect a variation in the renal handling of growth hormone during pubertal development. uGH estimation may be an unreliable screening investigation for growth hormone sufficiency in mid to late puberty.