The computerized dietary analysis system that was in use at Tufts University/Frances Stern Nutrition Center and two commercially available systems were compared in a study examining differences between systems. Content, software, data base maintenance, and cost were among the characteristics in which they differed. Calculations on three-day food records provided by five omnivore and five vegetarian women differed from one system to the next. Significant differences were noted for total kilocalories, total fat, saturated fatty acids, polyunsaturated fatty acids, cholesterol, and phosphorus. Unfortunately, it was not possible to describe the degree to which each of the many possible causes of variability may have contributed to the differences in nutrient calculations. However, coding activity did not appear to influence nutrient calculations within one system in a significant manner when skilled coders were employed. We conclude that differences observed in dietary studies using different computerized dietary analysis systems may be due in part not to real differences in subjects but rather to the system employed. Valid reference standards against which each nutrient value can be compared to ascertain exactly what these errors are do not exist. All the information consumers need to make useful comparisons among the various computerized dietary analysis systems is not yet available in a standardized format and is eagerly awaited by this and other groups.