Objectives: (1) To assess the test-retest reliability of physical performance tests in subject groups with different levels of impairment and disability, and (2) to assess the stability of these tests over different time intervals. Design: Test-retest, repeated measures reliability design. Setting: (1) A university's center for aging and research center, (2) a continuing care retirement community, and (3) an extended care and rehabilitation center at a Veterans Affairs medical center. Subjects: Twenty- four community-dwelling elders, 15 community-dwelling elders with Parkinson disease, 12 older women with vertebral osteoporosis and compression fractures, and 14 elderly nursing home residents. Measures: Lower extremity isometric strength (ankle dorsiflexion, hip abduction), spinal configuration (thoracic kyphosis, lumbar lordosis), lumbosacral motion (flexion, extension), and timed measures of the ability to get in and to get out of bed at a usual pace. Results: Most of the within-group intraclass correlation coefficients (ICCs) were good to excellent (.70 to .97). Overall, ICCs for all groups combined were between .70 and .96, and no decrement in reliability was noted after controlling for group membership. In addition, no decrement in the ICC was observed for short (1 day) vs longer (1 week) intervals of testing. Conclusions: These performance-based measures may be used reliably across a wider range of testing environments and elderly populations than has been reported.