RESPONDENT BIAS IN RETROSPECTIVE FOOD FREQUENCY QUESTIONNAIRES AND PROSPECTIVE DIETARY RECORDS

A possible contaminant of much dietary research - respondent bias - might result in people altering their eating habits or descriptions of their eating habits to what might seem to be more socially desirable. Data collected in a series of studies with 51 preschoolaged boys permitted two tests of respondent bias. In the first, semi-quantitative food frequency questionnaires for one month were used to assess food consumption patterns. The results were compared to weighed food records subsequently recorded for 21 days. In the food frequency questionnaires parents' estimates of their children's intake of a number of nutritious foods were significantly higher than the weighed food records. In the second test of respondent bias, three-day dietary records of dinners from each of three consecutive weeks were analyzed for nutrient content and compared to see if alterations in nutrient intake occurred during the study period. No differences in nutrient intake were found across the three consecutive weeks, suggesting that respondent bias may not affect dietary records. The consistency of the three sets of dietary records also provided support for the suggestion that three-day dietary records for this sample of children were representative of intake over 21 days.