A 60-item food sort patterned after a game used in anthropological studies was developed to measure food intake of 19 preschool children attending a day care center in Southwest Virginia. Response agreement was determined by administering the game twice to children one month apart, followed by phone interviews with parents, school menu review and teacher interview to confirm child response. Most three-year-olds completed only 20 out of 60 items; nine four- to five-year-olds were able to complete all 60. Children were fairly consistent in response over a month's time, with rough agreements 80 percent or higher for almost half the group; rough parent-child agreements were also reasonably good, with more than half scoring 75 percent or higher and none scoring below 60 percent. Dietary assessment was only possible to a limited degree, but food group frequency scores revealed children's estimates of their own intake to be considerably higher than those of parents. While reliability of children as primary sources of dietary data is suggested, further study is needed before true validity and reliability of the instrument can be established. / M.S.