Two hundred and nineteen second generation Japanese-American men were classified with a 75-g oral glucose tolerance test: 77 with normal glucose tolerance, 74 with impaired glucose tolerance (IGT), and 68 with noninsulin-dependent diabetes mellitus (NIDDM). The peak insulin response to the oral glucose load was progressively delayed with each of the 3 glucose tolerance categories. A similar finding was observed with the peak C-peptide response to oral glucose, except for the absence of distinction between IGT and NIDDM. Variables measuring the initial rate of insulin or C-peptide secretion (0-30 min) after oral glucose also demonstrated a progressive diminution with increasing glucose intolerance. The relative incremental insulin response at 30 min and the relative incremental C-peptide response at 30 min were highly correlated with the fasting glucose levels (r = -0.61 and r = -0.62; P less than 0.0001, respectively). Variables measuring the 0-30 min secretory response had high variances, whereas the variance for fasting glucose was low. Twelve men who were initially classified as IGT subsequently developed NIDDM. These 12 men had significantly higher fasting glucose levels at baseline than the remaining men who did not develop diabetes, but the 30 min secretory parameters after oral glucose, although lower in those who subsequently developed diabetes, were not significantly different at baseline. However, if fasting glucose is used as a surrogate measure of secretory response, these 12 men appear to have had an impairment of oral glucose-stimulated insulin secretion antedating the development of NIDDM. The inability of the secretory parameters to detect the abnormality may be due to a type II statistical error, which may be resolved by a larger sample size.