UNLABELLED In this study, we compare the extent to which seven available definitions of sarcopenia and two related definitions predict the rate of falling. Our results suggest that the definitions of Baumgartner and Cruz-Jentoft best predict the rate of falls among sarcopenic versus non-sarcopenic community-dwelling seniors. INTRODUCTION The purpose of the study is to compare the extent to which seven available definitions of sarcopenia and two related definitions predict the prospective rate of falling. METHODS We studied a cohort of 445 seniors (mean age 71 years, 45 % men) living in the community who were followed with a detailed fall assessment for 3 years. For comparing the rate of falls in sarcopenic versus non-sarcopenic individuals, we used multivariate Poisson regression analyses adjusting for gender and treatment (original intervention tested vitamin D plus calcium against placebo). Of the seven available definitions, three were based on low lean mass alone (Baumgartner, Delmonico 1 and 2) and four required both low muscle mass and decreased performance in a functional test (Fielding, Cruz-Jentoft, Morley, Muscaritoli). The two related definitions were based on low lean mass alone (Studenski 1) and low lean mass contributing to weakness (Studenski 2). RESULTS Among 445 participants, 231 fell, sustaining 514 falls over the 3-year follow-up. The prospective rate of falls in sarcopenic versus non-sarcopenic individuals was best predicted by the Baumgartner definition based on low lean mass alone (RR = 1.54; 95 % CI 1.09-2.18) with 11 % prevalence of sarcopenia and the Cruz-Jentoft definition based on low lean mass plus decreased functional performance (RR = 1.82; 95 % CI 1.24-2.69) with 7.1 % prevalence of sarcopenia. Consistently, fall rate was non-significantly higher in sarcopenic versus non-sarcopenic individuals based on the definitions of Delmonico 1, Fielding, and Morley. CONCLUSION Among the definitions investigated, the Baumgartner definition and the Cruz-Jentoft definition had the highest validity for predicting the rate of falls.