We report a cross-sectional study of impaired glucose tolerance conducted on data collected during the Fiji Cardiovascular and Metabolic Disease Survey, which was performed in 1980. Both Melanesian Fijians and migrants from the Indian subcontinent were studied. Impaired glucose tolerance and diabetes were defined according to current WHO criteria. When impaired glucose-tolerant subjects were compared with normals and diabetics with respect to selected physiological, biochemical and environmental factors, those with impaired glucose tolerance showed greater affinities with the diabetics than with the normal subjects. This finding was true of both the Melanesian and Indian populations, but was in contrast to a study of Micronesians in Kiribati (using standardized survey methods and similar analyses) in which subjects with impaired glucose tolerance held an intermediate position between normals and diabetics, with respect to the chosen factors. A possible explanation for the discrepant findings of the 2 studies is that the cut-off point in the distribution of plasma glucose concentration which defines diabetes may vary in populations of different ethnic origin, and the studies support the concept of heterogeneity within the category of impaired glucose tolerance.