Sources of variance in dietary intake records were measured and compared. Eighteen graduate students recorded self-selected food intakes during a 30-day study. Subjects were divided into two groups and on alternating 5 consecutive day periods, one group weighted all food consumed and recorded it on tape recorders while the other group recorded estimated intake in small notebooks. An average of two, 24-h dietary recalls was also obtained from each subject and compared to previously estimated intake. The major contributions to total variance were within subject and between subject variance. There was no difference in the mean protein and energy intake recorded by weighing when compared to that recorded in household measures, but there was less variability in the records obtained by weighting intake. The 95% confidence limits were calculated both for the group's and an individual's intake for each method, these demonstrated that 1-day dietary record gave a reasonable estimate (within 15%) of the usual intake of the group. However, the 1-day record, regardless of which method used, gave a meaningless estimate of an individual's usual diet. The protein/energy ratios and the coefficient of interindividual variation are also reported.