Mailed surveys may be a cost-effective way to collect dietary information in large populations. We sent a brief quantified food frequency survey to a random sample of 2,000 prepaid health plan members between the ages of 20 and 65 years for two consecutive years. Logistic regression was used to predict the probability of: (1) responding to the questionnaire, (2) completing all 60 items in the questionnaire correctly so that all information could be used (responding entirely correctly), and (3) completing all 60 items in the questionnaire incorrectly so that no information was usable (responding entirely incorrectly). In a follow-up survey of a subsample of respondents, Spearman correlation coefficients were used to compare nutrient intakes calculated from questionnaires that had omitted data with those from the same questionnaires after the missing information was completed by telephone interview. About 50% of all persons surveyed returned the questionnaire. Response rates were highest among persons who had taken a voluntary multiphasic health examination, among women, and among older persons. Only 22.9% of the respondents in year 1 and 27.7% of the respondents in year 2 responded entirely correctly; a small percentage of respondents (12.1%) omitted more than 10 food items. Responding entirely correctly was inversely related to age and related to race. Most correlations between nutrient intakes calculated from questionnaires with omitted data and those from the same questionnaire subsequently completed by telephone interview were greater than 0.90 and increased as the number of omitted items decreased. If a relatively low overall response rate is acceptable, food frequencies collected through the mail can be an effective way to gather dietary information.