Retroviral vector-mediated gene transfer has been central to the development of gene therapy. Retroviruses have several distinct advantages over other vectors, especially when permanent gene transfer is the preferred outcome. The most important advantage that retroviral vectors offer is their ability to transform their single stranded RNA genome into a double stranded DNA molecule that stably integrates into the target cell genome. This means that retroviral vectors can be used to permanently modify the host cell nuclear genome. Recently, retroviral vector-mediated gene transfer, as well as the broader gene therapy field, has been re-invigorated with the development of a new class of retroviral vectors which are derived from lentiviruses. These have the unique ability amongst retroviruses of being able to infect non-cycling cells. Vectors derived from lentiviruses have provided a quantum leap in technology and seemingly offer the means to achieve significant levels of gene transfer in vivo.The ability of retroviruses to integrate into the host cell chromosome also raises the possibility of insertional mutagenesis and oncogene activation. Both these phenomena are well known in the interactions of certain types of wild-type retroviruses with their hosts. However, until recently they had not been observed in replication defective retroviral vector-mediated gene transfer, either in animal models or in clinical trials. This has meant the potential disadvantages of retroviral mediated gene therapy have, until recently, been seen as largely, if not entirely, hypothetical. The recent clinical trial of gammac mediated gene therapy for X-linked severe combined immunodeficiency (X-SCID) has proven the potential of retroviral mediated gene transfer for the treatment of inherited metabolic disease. However, it has also illustrated the potential dangers involved, with 2 out of 10 patients developing T cell leukemia as a consequence of the treatment. A considered review of retroviral induced pathogenesis suggests these events were qualitatively, if not quantitatively, predictable. In addition, it is clear that the probability of such events can be greatly reduced by relatively simple vector modifications, such as the use of self-inactivating vectors and vectors derived from non-oncogenic retroviruses. However, these approaches remain to be fully developed and validated. This review also suggests that, in all likelihood, there are no other major retroviral pathogenetic mechanisms that are of general relevance to replication defective retroviral vectors. These are important conclusions as they suggest that, by careful design and engineering of retroviral vectors, we can continue to use this gene transfer technology with confidence.