Objective: To study prospectively the association of coffee intake with incident diabetes in the Puerto Rico Heart Health Program cohort, comprising 9824 middle-aged men (aged 35-79 years). Methods: Of 9824 men, 3869 did not provide a fasting blood sample at baseline, 1095 had prevalent diabetes and 131 were not given fasting glucose tests at any subsequent study visit. Thus, the present analysis includes 4685 participants. Diabetes was ascertained at baseline and at two study visits between 1968 and 1975 using fasting glucose tests and self-reports of physician-diagnosed diabetes or use of insulin or hypoglycaemic medication. Logistic regression analysis was used to assess the association of coffee intake with risk of incident diabetes while adjusting for covariates (age, BMI, physical activity, smoking, education, alcohol intake, family history of diabetes, intakes of milk and sugar). Results: Five hundred and nineteen participants met the criteria for incident diabetes. Compared with those reporting intake of 1?2 servings of coffee/d, coffee abstainers were at reduced risk (OR = 0·64; 95 % CI 0·43, 0·94). Among coffee drinkers, there was a significant trend of decreasing risk by intake (P = 0·02); intake of ≥4 servings/d was associated with an odds ratio of 0·75 (95 % CI 0·58, 0·97).. Conclusions: Study findings support a protective effect of coffee intake on diabetes risk, while also suggesting that abstainers may be at reduced risk.