The accuracy of the Minnesota Leisure Time Physical Activity (LTPA) Questionnaire (a 1-year survey of non-occupational activity used in MRFIT) was studied in 78 men and women, age 20-59 years, by comparing survey results to the following measures obtained over a year's duration: six 48-hour physical activity records; fourteen 48-hour Caltrac accelerometer readings (Caltrac); 14 administrations of a 4-week version of the LTPA Questionnaire (FWH); 3 VO2peak determinations and percent body fat (% BF). The LTPA Questionnaire demonstrated: weak to moderate associations with % BF (r = -0.24) and with VO2peak (r = 0.47); a weak association with Caltrac readings expressed as MET.minutes.day-1 (r = 0.23); strong associations with corresponding activities reported on the FWH; and moderate associations with total and heavy activities reported in the physical activity record, but no associations with moderate and light activities. Furthermore, several types of LTPA were found to be either under-represented or not currently included in the Minnesota LTPA Questionnaire. It is concluded that although validation results were found to be quite good, several possible refinements were identified, which should improve the accuracy of the Minnesota LTPA Questionnaire in assessing habitual physical activity.