Aims/hypothesis. Our studies were undertaken to characterise the defective insulin secretion of impaired glucose tolerance (IGT).¶Methods. We studied 13 normal glucose tolerant subjects (NGT) and 12 subjects with IGT carefully matched for age, sex, BMI and waist-to-hip ratio. A modified hyperglycaemic clamp (10 mmol/l) with a standard 2-h square-wave hyperglycaemia, an additional glucagon-like-peptide (GLP)-1 phase (1.5 pmol · kg^–1· min^–1 over 80 min) and a final arginine bolus (5 g) was used to assess various phases of insulin secretion rate.¶Results. Insulin sensitivity during the second phase of the hyperglycaemic clamp was low in both groups but not significantly different (0.12 ± 0.021 in NGT vs 0.11 ± 0.013 μmol · kg^–1· min^–1· pmol^–1 in IGT, p = 0.61). First-phase insulin secretion was lower in IGT (1467 ± 252 vs 3198 ± 527 pmol · min^–1, p = 0.008) whereas the second phase was not (677 ± 61 vs 878 ± 117 pmol · min^–1, p = 0.15). The acute insulin secretory peak in response to GLP-1 was absent in IGT subjects who only produced a late phase of GLP-1-induced insulin secretion rate which was lower (2228 ± 188 pmol · min^–1) than in NGT subjects (3056 ± 327 pmol · min^–1, p = 0.043). Insulin secretion in response to arginine was considerably although not significantly lower in IGT subjects. The relative impairment (per cent of the mean rate for NGT subjects) was greatest for the GLP-1 peak (19 ± 9 %).¶Conclusion/interpretation. In this Caucasian cohort a defective insulin secretion rate is essential for the development of IGT. The variable degrees of impairment of different phases of the insulin secretion rate indicate that several defects contribute to its abnormality in IGT. Defects in the incretin signalling pathway of the beta cell could contribute to the pathogenesis of beta-cell dysfunction of IGT and thus Type II (non-insulin-dependent) diabetes mellitus. [Diabetologia (2000) 43: 852–858]