Objective To develop and validate a consensus set of retrieval categories for how children remember what they have eaten, and to relate retrieval categories to accuracy of reporting items eaten during school lunch.

Design A Delphi technique study was conducted using 10 psychologists. The subject matter consisted of responses transcribed from interviews conducted with 89 randomly selected fourth graders within 90 minutes of eating. Retrieval categories were evaluated for accuracy by comparing students’ self-reported lunch intake with observation.

Statistical analyses For round 1, a centroid hierarchical cluster analysis was used to identify common sets of rating pairs to propose categories for reaching consensus. For rounds 2 and 3, percent of agreement was calculated. Accuracy was tabulated across meal items and student use by retrieval category.

Results After round 3, we found that 23 “near-consensus”categories were used by at least eight raters. Six categories were used 60% of the time, 4 were used 20% of the time, and 5 were used 15% of the time. Less frequently used categories were combined with similar, more frequently used categories for a total of 16 categories. Students used a large variety of retrieval categories when accurately reporting consumption. Of the 16 categories, 12 were used similarly by both accurate and inaccurate students. Where there were differences in accuracy by retrieval category, the more accurate students used “taste/smell/texture”and “visual”in deference to “order items consumed”and “oral cue still present.”

Applications These categories provide insight into the broad range of retrieval categories that children use and provide direction for researchers to design and study specific cues to enhance the accuracy of children's self-reports of diet. J Am Diet Assoc. 1997;97:31–36.