To compare two methods for classifying an individual as sarcopenic for predicting decline in physical function in the Health, Aging and Body Composition Study. Observational cohort study with 5 years of follow-up. Communities in Memphis, Tennessee, and Pittsburgh, Pennsylvania. Men and women aged 70 to 79 (N=2,976, 52% women, 41% black). Appendicular lean mass (aLM) was measured using dual energy x-ray absorptiometry, and participants were classified as sarcopenic first using aLM divided by height squared and then using aLM adjusted for height and body fat mass (residuals). Incidence of persistent lower extremity limitation (PLL) was measured according to self-report, and change in objective lower extremity performance (LEP) measures were observed using the Short Physical Performance Battery. There was a greater risk of incident PLL in women who were sarcopenic using the residuals sarcopenia method than in women who were not sarcopenic (hazard ratio (HR)=1.34, 95% confidence interval (CI)=1.11-1.61) but not in men. Those defined as sarcopenic using the aLM/ht(2) method had lower incident PLL than nonsarcopenic men (HR=0.76, 95% CI=0.60-0.96) and women (HR=0.75, 95% CI=0.60-0.93), but these were no longer significant with adjustment for body fat mass. Using the residuals method, there were significantly poorer LEP scores in sarcopenic men and women at baseline and Year 6 and greater 5-year decline, whereas sarcopenic men defined using the aLM/ht(2) method had lower 5-year decline. Additional adjustment for fat mass attenuated this protective effect. These findings suggest that sarcopenia defined using the residuals method, a method that considers height and fat mass together, is better for predicting disability in an individual than the aLM/ht(2) method, because it considers fat as part of the definition.