ISEE-0307 Background and Objective: There is increasing interest in the impact of the built nutritional environment on health behaviors and outcomes. Little is known about the validity of readily available databases from public or commercial sources. Methods: We conducted a comprehensive field census of the rural food environment in seven South Carolina counties identifying all food outlets (i.e. supermarkets, grocery stores, food wholesalers, discount stores, drug stores, restaurants, delis and specialty stores, convenience stores with or without gas stations). Locations were acquired using a handheld global positioning system unit (GPS, Trimble Juno ST). A data source from a state agency (DB-A) and a commercial business listing (DB-B) were compared to field census results. We computed concordance (number agree/total number identified in field census or database), sensitivity (number agree/number in field census), and positional accuracy (Euclidian distance between GPS and geocode among concordant outlets). Results: The field census identified a total of 1,054 food outlets, DB-A listed 823 outlets and DB-B listed 767 outlets. DB-A and the field census showed good agreement (concordance = 0.63) on the presence of food outlets, while DB-B showed only moderate agreement (concordance = 0.47). The sensitivity was quite good for DB-A at 69% but poor for DB-B at 53%. The positional accuracy varied tremendously, with a median difference of 122 meters for DB-A and 104 meters for DB-B. Despite these spatial discrepancies, 77.3% of food outlets in DB-A and 79.8% of those in DB-B were allocated to the correct census tract. Conclusion: Our study suggests that in rural areas the validity of readily available data sources used for the characterization of the nutritional environment differs markedly in terms of completeness and spatial accuracy. Both undercount of food outlets and positional inaccuracies can introduce bias into studies evaluating the impact of the built nutritional environment on health.