Food-frequency questionnaires are usually administered as a list of foods to be checked off by the respondent or interviewer. Techniques in which participants sort into categories cards on which names or pictures of foods are printed can also be used to assess food intake. Food-frequency scores were obtained from a five-category picture sort administered to 4643 men and women aged > or = 65 y in the Cardiovascular Health Study (CHS). This one-step (qualitative) assessment yielded significant associations in expected directions between frequency scores and sex, age, race or ethnicity, body mass index, and use of a special diet. In the two-step (semiquantitative) version of this instrument, an interviewer documented specific frequencies and portion-size information for the foods in each sorting category. A substudy of the two-step version with 96 CHS participants indicated relative validity similar to that of conventionally administered food-frequency questionnaires. The one-step version may be broadly applicable to situations in which general food-pattern data can be informative and cost and time limitations are great. When it is feasible, the two-step picture sort may offer certain methodologic advantages because respondents have a chance to change their responses and the format may simplify the cognitive-response task. Sorting or picture-sort procedures deserve systematic attention in research on dietary assessment methods.