Research suggests an association between hyperkyphosis and falls in community-dwelling older adults, though this has not been investigated within large, population-based studies. This study sought to determine whether two measures of kyphosis prospectively predict fall risk over 3years among older men. Within the Osteoporotic Fractures in Men Study (MrOS), we conducted two 3-year prospective studies of 2346 and 2928 men. The first group had kyphosis measured by the Cobb angle at visit 1, while the second group had kyphosis assessed with the blocks method at visit 3; both groups then self-reported falls tri-annually for 3years. Poisson regression with GEE was used to obtain relative risks (RR) of falls. The fall rates over 3years were 651/1000 person-years among the visit 1 sample (mean age 74 ± 6years) and 839/1000 person-years among the visit 3 sample (mean age 79 ± 5years). In adjusted models of the visit 3 sample, the risk of falls was increased by 12% for each standard deviation increase (1.4 blocks) in the number of blocks required to achieve a neutral head and neck position (RR = 1.12, 95% CI = 1.06, 1.18). The Cobb angle was not associated with falls in the visit 1 sample. Although the Cobb angle did not predict falls in community-dwelling older men over 3years, the blocks method of measuring kyphosis was predictive of falls in this population. This difference could be due to the Cobb angle's focus on thoracic kyphosis, whereas the blocks method may additionally capture abnormal cervical spine curvature.