In view of the important role that nutrient intake assessments play in establishing the Recommended Dietary Allowances (RDAs), the quality of food composition data must be assured for accuracy and representativeness. Assurance of data quality requires the definition of critical parameters in the data generation process and the evaluation of specific data for foods and components according to these parameters. An expert systems approach for evaluating the quality of analytical data has been developed by scientists at the Beltsville Human Nutrition Research Center to determine the quality of food composition data for five parameters: sampling plan, sample handling, number of samples, analytical method and analytical quality control. A rating scale for each parameter was developed with 0 representing poor or inadequately documented data and 3 representing optimal data. Specific criteria for each parameter and rating have been developed and incorporated into expert systems software to facilitate the objective assignment of ratings for each data source by the reviewer. After all ratings for a specific food-nutrient combination are assigned, the system calculates a composite score called the "confidence code" which indicates to the user the relative level of confidence in the data. By identifying and rating the important steps in the data generation process, one can begin to partition the possible sources of error or variability in the process. Limitation of the data set relative to specific purposes (e.g., setting the RDAs) can be identified. The evaluation process can provide the basis for focussed research to improve the most critical areas of the data generation process. A similar process could be established to evaluate the quality of analytical data for clinical measurements used to establish the RDAs.