Although achievement of many of the Healthy People Year 2000 nutrition objectives will be determined by dietary assessment, methods for assessing diet are weak, especially for children. Children's self-reports of diet are necessary because parents do not know what children eat at school. Self-reports rely on memory, which is subject to error. Research concerning memory and dietary assessment is limited. The purpose of this study was to compare how 4th-graders remember items and amounts eaten at the school lunch the same day with how other 4th-graders remember items and amounts eaten at the school lunch the previous day. Students from 1 6 classes at 3 schools were randomly selected, observed eating lunch, and interviewed within 2 hours after eating (n = 89) or the following morning (n= 16). Interviews were audio-recorded and included free report, cognitive aspects of free report, prompted report, and cognitive aspects of prompted report. For analyses, responses obtained during each of the two cognitive phases (free, prompted) from same day interviews were grouped according to item, detail, and amount, and categorized into 14 categories of responses (preferences, smell/taste imagery, texture imagery, visual imagery, usual practice, food-related events, read label, read menu, behavior chaining, amount eaten, weight of container, physiological cues, other, unable to verbalize). The process was repeated for next day reports. Chi-square analyses were conducted with the individual student as the unit of analysis. Usual practice was more commonly used during next day than same day recalls for items freely reported (81 % vs. 40%, p<.003), and for items reported after prompting (25% vs. 0%, p<.01). Preferences was reported by more students during next day than same day recalls for amounts freely reported (73% vs. 39%, p<.04). Food-related events (e.g., spill or stain on clothes or body) was more commonly used during next day than same day recalls for details freely reported (14% vs. 0%, p<.02). Texture imagery was more common during next day than same day recalls for details reported after prompting (15% vs. 3%, p<.05). Read the menu was more commonly used during next day than same day recalls for details reported after prompting (31 % vs. 10%, p<.04). The inability to verbalize how the information was remembered was more common during same day than next day recalls for items freely recalled (24% vs. 0%, p<.03). Enhanced understanding of how children remember what they have eaten may aid in the design of cues to improve the accuracy of children's self-reports of diet. More accurate dietary self-reports could facilitate assessment of the Year 2000 nutrition objectives.