Grafting neural stem cells is a widely used experimental approach to central nervous system (CNS) repair after trauma or neurodegeneration. It is likely to be a realistic clinical therapy for human CNS disorders in the near future. One of the challenges of this approach is the ability to identify both the survival and differentiated phenotype of various stem cell populations after engraftment into the CNS. There is no single protocol that will work for all cell types and all applications. Labeling stem cells for CNS grafting is an empirical process. The type of stem cell, its fate after engraftment, and the context in which it is anatomically and histologically evaluated all contribute to a decision as to the best approach to take. We have provided the range of conditions under which various labels have been successfully used in CNS grafting studies and delineated the parameters that have to be empirically established. Given a clear understanding of the limitations of the respective labels and the expected outcome of the grafting experiment, these labeling guidelines should enable any investigator to develop a successful approach. Our own personal bias is to use labels that cannot be transferred to host cells. Initially, we preferred 5-bromo-2′-deoxyuridine, or retrovirally delivered enhanced green fluorescent protein or lacZ. More recently, we have found syngeneic grafts of human placental alkaline phosphatase stem cells to work very well. However, each investigator will have to decide what is optimal for his or her cell population and experimental design. We summarize the various approaches to labeling and identifying stem cells, pointing out both the limitations and strengths of the various approaches delineated.