Gene addition strategies are rational approaches to the treatment of sickle cell anemia and thalassemia. The goal of such genetic treatments is to introduce a functional globin transcription unit in hematopoietic stem cells and express the transgene in a manner that is erythroid-specific, elevated, relatively constant from one cell to another, and sustained over time. Gene transfer is mediated by an expanding array of viral and nonviral vectors. High-titer retroviral vectors harboring the human beta-globin gene and the core sequences of the human beta-globin locus control region yield erythroid-specific gene expression in erythroid cell lines and in short-term murine bone marrow chimeras. However, we show that expression remains subject to position effect variegation and often decreases over time in vivo. Rather than a progressive transcriptional silencing in all cells, we ascribe the waning expression to the gradual emergence in blood of erythroid progeny derived from more and more primitive precursor cells in the months after transplantation. In our model, transgene expression is therefore determined by the integration site and the differentiation stage of the transduced cell at the time of integration. Globin expression is thus different in the progeny of a transduced erythroid progenitor cell and in the erythroid progeny of a transduced hematopoietic stem cell, reflecting the effect of flanking chromatin in differentiated cells and of chromatin remodeling at the site of integration in the progeny of multipotential cells. This model predicts that insulators and matrix attachment regions could be highly valuable to gene therapy in combination with potent transcriptional activators. When efficient gene transfer in hematopoietic stem cells is achieved at last, the challenge will be to regulate gene expression in vivo and overcome transgene variegation and transgene silencing.