The aims of this study were to investigate the types of error made after intervals of 1 to 8 days in reporting foods recorded as eaten over a 24-hour period and to examine ways of improving memory for foods eaten. For Study 1, male and female university sta and student volunteers were instructed to keep a record of all dietary intake for a 24-hour period and to return after 1, 2, 4 or 8 days to report orally the recorded day's intake using a standard protocol for dietary recall. In Study 2, 67 similar volunteers returned at 4 days after recording and orally reported foods eaten using the standard protocol or one of three protocols with added prompts used in a `cognitive interview', namely context reinstatement, repeating the recall or recalling again in the reverse temporal sequence. Oral report of food items was compared with the written record for the number of correct food items and errors of omission, intrusion, description and quantity as a ratio of the recorded items. The number of foods remembered correctly declined with increasing time after recording. Omissions of food items increased after 4 and 8 days and may be attributed to confusion from intervening days. In contrast, intrusion errors were made at all recall intervals, indicating confusion between the recorded day's intake and usual diet. This type of error could make an otherwise correct report in the 24-hour recall approach to dietary assessment more representative of the individual's usual intake. Neither of the repeat recall procedures reduced errors. However, reinstatement of the context of eating occasions may have potential for improving the accuracy of reports of foods eaten in the past, as in dietary assessments that rely on memory.