In a retrospective analysis, we have compared the response of serum GH concentration to insulin-induced hypoglycaemia in 148 short prepubertal children (114 males, 34 females) aged between 3·9 and 11·9 years with the growth rate of the individual to determine 'cut-off' values for the diagnosis of GH insufficiency. Sixty-three children grew with a height velocity standard deviation score (SDS) greater than −0·8 (group 1), which represents the growth velocity of children progressing along or closely parallel to the third height centile. Eighty-five children had a height velocity SDS of less than −0·8 (group 2). Median peak serum GH concentration responses to insulin-induced hypoglycaemia were 19·9 mU/l (range 1·5–54·4) in group 1 and 9·9 mU/l (range 0·7–46·2) in group 2 (Mann–Whitney; P < 0·001). Using growth rate as the determinant of normality, the efficiency, sensitivity and specificity of the insulin-induced hypoglycaemia test were calculated using different serum GH concentration cut-off values to diagnose GH insufficiency. In our (Hybritech) assay, a cut-off value of 13·5 mU/l provided optimal performance in terms of efficiency (66%), sensitivity (64%) and specificity (70%). The response of serum GH concentration to insulin-induced hypoglycaemia in short children growing at different growth rates was continuous. Each laboratory measuring serum GH concentrations needs to construct its own 'normal' cut-off value. Journal of Endocrinology (1992) 133, 447–450