Retinol is involved in bone remodeling, and excessive intake has been linked to bone demineralization, yet its role in osteoporosis has received little evaluation. We studied the associations of retinol intake with bone mineral density (BMD) and bone maintenance in an ambulatory community-dwelling cohort of 570 women and 388 men, aged 55-92 years at baseline. Regression analyses, adjusted for standard osteoporosis covariates, showed an inverse U-shaped association of retinol, assessed by food-frequency questionnaires in 1988-1992, with baseline BMD, BMD measured 4 years later, and BMD change. Supplemental retinol use, reported by 50% of women and 39% of men, was an effect modifier in women; the associations of log retinol with BMD and BMD change were negative for supplement users and positive for nonusers at the hip, femoral neck, and spine. At the femoral neck, for every unit increase in log retinol intake, supplement users had 0.02 g/cm2 (p = 0.02) lower BMD and 0.23% (p = 0.05) greater annual bone loss, and nonusers had 0.02 g/cm2 (p = 0.04) greater BMD and 0.22% (p = 0.19) greater bone retention. However, among supplement users, retinol from dietary and supplement sources had similar associations with BMD, suggesting total intake is more important than source. In both sexes, increasing retinol became negatively associated with skeletal health at intakes not far beyond the recommended daily allowance (RDA), intakes reached predominately by supplement users. This study suggests there is a delicate balance between ensuring that the elderly consume sufficient vitamin A and simultaneously cautioning against excessive retinol supplementation.