The objective of this study was to determine the longitudinal prevalence and predictors of dietary underreporting in African-American preadolescent girls and the association of baseline dietary underreporting with changes in BMI over a 2-year period as part of the Girls health Enrichment Multi-site Studies (GEMS). Energy was summarized at baseline, 12 months, and 24 months and computed as a 3-day average of 24-hour dietary recalls. Physical activity was assessed by accelerometer, basal metabolic rate was estimated using the World Health Organization's prediction equation, and caloric underreporting was based on the Goldberg equation. We classified 48% of the girls at baseline as underreporters; with underreporting increasing over time (61% at 12 months; 66% at 24 months). Intervention treatment assignment did not affect the prevalence of underreporting over time. The consistency of underreporting (or not) stayed stable over time. Across all three time points, a higher BMI predicted underreporting. Baseline dietary underreporting and baseline BMI were found to be the major predictors of change in BMI, whereas baseline dietary variables did not predict change in BMI. Dietary underreporting was extremely common in this sample of AA preadolescent girls and predictive of change in BMI. Given the magnitude and consistency of dietary underreporting along with the fact that no dietary variables predicted change in BMI, measurement of dietary intake in preadolescents, even with sophisticated measurement methodologies, appears biased. The best use of dietary recalls may not be to estimate dietary intake but rather to determine underreporting.