Reliability of recall of physical activity in the distant past

Am J Epidemiol. 1999 Jul 15;150(2):195-205. doi: 10.1093/oxfordjournals.aje.a009980.


Substantial data exist supporting the role of physical activity in the etiology of several chronic diseases. Many chronic diseases begin developing 20-30 years before they become clinically evident. Since researchers often must rely on recall to characterize the long term habits of study participants, the accuracy of recall of physical activity is an important methodological issue in etiologic studies. The purpose of this study was to examine the quality of recall of physical activity in the distant past in a cohort of western New York residents followed since 1960. Paired t tests and intraclass correlation coefficients (ICCs) were used to compare "original" (1960) and "recalled" (1992-1996) reports of weekday (occupational) and free-day (leisure time) physical activity. Results showed that the recalled reports underestimated past weekday activities when overall activity was examined; estimates closer to the originals were found when levels of activity were examined. Recall was best for weekday light (ICC = 0.43) and weekday moderate (ICC = 0.45) activity in both sexes and free-day hard activity in females (ICC = 0.45). Most participants underestimated past free-day activity, but males overestimated free-day hard activity. Correlations for free-day activity were highest for summer sports in females (ICC = 0.29) and winter sports in both sexes (ICC = 0.39) and were low for walking and "other activity." Considering the length of time between the original interviews and the recall interviews, the correlations found here are remarkable and close to those found in other studies where recall intervals were 10 years or less.

Publication types

  • Research Support, U.S. Gov't, P.H.S.

MeSH terms

  • Adult
  • Cohort Studies
  • Epidemiologic Methods
  • Exercise
  • Female
  • Humans
  • Male
  • Mental Recall*
  • New York / epidemiology
  • Physical Exertion*
  • Reproducibility of Results