This study examined the reproducibility and comparability of five measures of function and exercise tolerance. The test battery and questionnaire on function and physical activity were administered twice, 7-10 days apart to 38 men and 12 women aged 54-80 years at the Baltimore Veterans Affairs Medical Center. Tests included fast pace 4 and 20-meter walks, 6-minute and graded treadmill walks, and a seated step test. All tests demonstrated good reproducibility with Pearson and intraclass correlation coefficients ranging from 0.84 to 0.98, and percent differences on retest ranging from 4 to 11%. Although correlations between different tests were all significant (range 0.34-0.89), comparison of performance ranks and linear regression analyses indicated that the short fast walks and seated step test may not be suitable substitutes for treadmill or long self-paced corridor walks. Only 28% had the same quintile performance ranking on the step test as on the treadmill walk, and 36% had rankings 2 or more points apart. The fast 20m walk shows the most promise as a low-level alternative to the 6-minute walk; performances had a correlation of 0.73, 82% of ranks were within one point, and 20m speed explained 42% of the variance in distance covered. More development is needed for comprehensive assessment of exercise tolerance in older adults; the 6-minute walk did not adequately discriminate fitness level in persons who walk regularly, and the treadmill posed problems for those with walking difficulty.