There is essentially no ideal method of dietary assessment. Physiologic methods (ie, doubly labeled water) probably come closest, but high cost, participant burden, and limited information limit their use. Furthermore, most dietary assessment methods have been designed for and tested in white adults. Very few have been designed for and tested in African-American adolescents. This study examined validity and reliability of a modified picture-sort food frequency questionnaire (FFQ) administered to 22 low-income, overweight, African-American adolescent girls, aged 11 to 17 years. The FFQ was administered to subjects twice during a 2-week period, and evaluated using the mean values of three 24-hour recalls. The natural log-transformed energy-adjusted, deattenuated correlation coefficients between the second FFQ and the mean from 3 recalls exceeded 0.50 for most nutrients, ranging from 0.32 (protein) to 0.87 (saturated fat). The energy and nutrient values from the first FFQ were greater than those from the second FFQ. Most correlation coefficients for the test-retest reliability of the FFQ were not significant. We conclude that although larger samples are needed to generalize results, the picture-sort dietary assessment method appears to be promising and merits further research. J Am Diet Assoc. 2000; 100:1050–1056