Embryonic stem cells (ESCs) hold great therapeutic potential due to their ability to differentiate into cells of the three primary germ layers, which can be used to repopulate disease-damaged tissues. In fact, two cell therapies using ESC derivatives are currently in phase I clinical trials. A main concern in using ESCs and their derivatives for cell transplantation is the ability of undifferentiated ESCs to generate tumors in the host. Positive selection steps are often included in protocols designed to generate particular cell types from ESCs; however, the transition from ESC to progenitor cell or terminally differentiated cell is not synchronous, and residual undifferentiated cells often remain. In our transplants of ESC-derived neural progenitors (ESNPs) into the adult mouse hippocampus, we have observed the formation of teratocarcinomas. We set out to reduce teratocarcinoma formation by enrichment of ESNPs using fluorescence-activated cell sorting (FACS) and have found that, although enrichment prior to transplant reduces the overall rate of teratocarcinoma formation, the tumorigenicity of cell batches can vary widely, even after FACS enrichment to as much as 95% ESNPs. Our data suggest that this variability may be due to the percentage of residual ESCs remaining in the transplant cell population and to the presence of pluripotent epiblast-like cells, not previously identified in transplant batches. Our data emphasize the need for stringent characterization of transplant cell populations that will be used for cell replacement therapies in order to reduce the risk of tumor formation.