A major strength of Arabidopsis thaliana as a model lies in the availability of a large number of naturally occurring inbred lines. Recent studies of A. thaliana population structure, using thousands of accessions from stock center and natural collections, have revealed a robust pattern of isolation by distance at several spatial scales, such that genetically identical individuals are generally found close to each other. However, some individual accessions deviate from this pattern. While some of these may be the products of rare long-distance dispersal events, many deviations may be the result of mis-identification, in the sense that the data regarding location of origin data are incorrect. Here, we aim to identify such discrepancies. Of the 5965 accessions examined, we conclude that 286 deserve special attention as being potentially mis-identified. We describe these suspicious accessions and their possible origins, and advise caution with regard to their use in experiments in which accurate information on geographic origin is important. Finally, we discuss possibilities for maintaining the integrity of stock lines.