We assessed the reproducibility and validity of a self-administered physical activity questionnaire used in a prospective study of 51,529 men. The questionnaire was administered by mail twice to 238 participants 2 years apart. During this interval, the participants completed a past-week recall and a 1-week activity diary at four times corresponding to different seasons throughout a year. Also, a step test was taken by a subset of the study participants. The intraclass correlation coefficients used to measure reproducibility were 0.39 for inactivity, 0.42 for nonvigorous activity, and 0.52 for vigorous activity. The correlations between diary-based and questionnaire-based activity scores, adjusted for variation in the diary measurements, were 0.41 for inactivity, 0.28 for nonvigorous activity, and 0.58 for vigorous activity. The distribution of activity scores was similar between the questionnaires and the average of past-week recalls, indicating the participants' ability to incorporate seasonal variation into their recall on a questionnaire. The correlation between vigorous activity and resting pulse was -0.45, whereas, for the pulse rate after stepping, the correlation was -0.41. These data indicate that this physical activity questionnaire is reproducible and provides a useful measure of average weekly activity, particularly vigorous activity, over a 1-year period.