The correlation between a releasable pituitary growth hormone (GH) pool and degree of growth failure was examined in 30 children with GH deficiency (group I) and 19 children with normal short stature (group II). Based on the responsiveness of GH to GH-releasing hormone (GHRH), group I, with low GH responses (below 7 ng/ml) to both insulin and arginine, was classified into three subgroups; Ia (peak value less than 10 ng/ml, n = 19), Ib (10-20 ng/ml, n = 5) and Ic (above 20 ng/ml, n = 6). Group II, with a GH response above 10 ng/ml to either insulin or arginine, was also divided into IIa (below 20ng/ml, n = 5) and Ifb (above 20ng/ml, n = 14). Body length and growth velocity in Ia and Ib were significantly reduced vs Ic; bone age in Ia was retarded vs L; plasma somatomedin C (Sm-C) levels in L and Ib were decreased vs Io, who had almost normal levels (0.90 _+ 0.55 U/ ml). The incidence of other combined pituitary hormone deficiencies and previous perinatal distress was definitely high in I~ and Ib, but zero in Ic. In group II also, body length and growth velocity were significantly decreased in II, vs IIb (P < 0.01). These results indicate that [1] the pituitary reserve of GH estimated by GHRH is a good reflection of the degree of growth failure in GH-deficient children as well as in those of normal short stature, [2] hypothalamic GHRH deficiency tends to have a milder effect on growth retardation than pituitary GH deficiency, and [3] normal short children with a diminished GH reserve may be potential candidates for the GH treatment.