Reduced vagal activity assessed by heart rate variability (HRV) has been observed in studies of diabetics, but this association has not been reported at the population level. To investigate the association of HRV with diabetes mellitus, as well as fasting serum insulin, and glucose, we examined a stratified random sample of 1933 individuals (154 diabetics and 1779 non-diabetics), aged 45-65 years from the Atherosclerosis Risk in Communities (ARIC) study cohort. Two-minute, resting, supine beat-to-beat heart rate records were collected. Power spectral density estimation was used to derive HRV high frequency power (HF, 0.15-0.35 Hz) as the conventional marker of vagal function. Age, race, and gender-adjusted geometric means of HF were 0.78 and 1.27 (beat/min)(2) for diabetics and non-diabetics respectively (P for mean difference <0.01), reflecting a reduced vagal activity in diabetics. In individuals not diagnosed as diabetics, a graded, inverse association was observed between fasting serum insulin and HF (P for trend <0.01): the age, race, and gender-adjusted geometric mean values of HF in the lowest and highest quartiles of serum insulin were 1.34 and 1.14 (beat/minute)(2), respectively. A similar association was observed between glucose and HF in a univariate model, but not in the adjusted model. This first population-based study on this subject confirmed that diabetics have significantly lower vagal activity than non-diabetics. In individuals not diagnosed as diabetics, serum insulin, and, to a lesser degree, serum glucose were inversely associated with vagal function, suggesting a role in the pathogenesis of diabetic neuropathy.