OBJECTIVE To determine the internal validity of the physical activity index (PAI). METHODS Three time periods over a 26-year period (1965/68 examination, n = 7911; 1980/82 examination, n = 2054; 1991/93 examination, n = 3425). METHODS Honolulu, Hawaii. METHODS Middle-aged to older Japanese-American men. METHODS PAI, other physical activity measures, body mass index (BMI), tricep skinfold thickness, subscapular skinfold thickness, and HDL cholesterol. RESULTS PAI levels was associated significantly with levels of other physical activity measures at all time periods. Higher PAI level was associated with lower BMI at the baseline examination (1965/68), lower subscapular skinfold thickness at the 1965/68 and 1980/82 examinations, lower tricep skinfold thickness at all examinations, and higher HDL cholesterol at the 1980/82 examination. Change in PAI level from 1965/68 to 1980/82 was associated with change in BMI, subscapular, and tricep skinfold thickness levels, but there were no significant associations between change in PAI level and change in BMI, skinfold thicknesses, or HDL cholesterol levels from 1980/82 to 1991/93. CONCLUSIONS PAI level is a relatively valid estimate of overall daily energy expenditure that may be an acceptable general measure across the middle-to-late stages of life. The PAI may also be sensitive enough to discern a shift toward less physical demanding activities as a population becomes older, although further studies should be conducted to better determine its validity in older populations.