OBJECTIVE: Coffee consumption is associated with reduced risk of type 2 diabetes, but the mechanism is not clearly understood. Elevated C-peptide, as a marker of insulin secretion, has been linked to insulin-resistant type 2 diabetes. In this study, we examined consumption of caffeinated and decaffeinated coffee and total caffeine in relation to concentrations of plasma C-peptide. RESEARCH DESIGN AND METHODS: Plasma C-peptide concentrations were measured in a cross-sectional setting among 2,112 healthy women from the Nurses' Health Study I who provided blood samples in 1989-1990. Consumption of caffeinated and decaffeinated coffee and total caffeine was assessed using a semiquantitative food-frequency questionnaire in 1990. RESULTS: Intakes of caffeinated and decaffeinated coffee and caffeine in 1990 were each inversely associated with C-peptide concentration in age-adjusted, BMI-adjusted, and multivariable-adjusted analyses. In multivariable analysis, concentrations of C-peptide were 16% less in women who drank >4 cups/day of caffeinated or decaffeinated coffee compared with nondrinkers (P < 0.005 for each). Women in the highest quintile compared with the lowest quintile of caffeine intake had 10% lower C-peptide levels (P = 0.02). We did not find any association between tea and C-peptide. The inverse association between caffeinated coffee and C-peptide was considerably stronger in obese (27% reduction) and overweight women (20% reduction) than in normal weight women (11% reduction) (P = 0.005). CONCLUSIONS: Our findings suggest a potential reduction of insulin secretion by coffee in women. This reduction may be related to other components in coffee rather than caffeine.