Paroxysmal nocturnal hemoglobinuria (PNH) results from somatic mutations in the PIG-A gene, leading to poor presentation of glycosylphosphatidylinositol (GPI)-anchored surface proteins. PNH frequently occurs in association with suppressed hematopoiesis, including frank aplastic anemia (AA). The relationship between GPI-anchored protein expression and bone marrow (BM) failure is unknown. To assess the hematopoietic defect in PNH, the numbers of CD34⁺ cells, committed progenitors (primary colony-forming cells [CFCs]), and long-term culture-initiating cells (LTC-ICs; a stem cell surrogate) were measured in BM and peripheral blood (PB) of patients with PNH/AA syndrome or patients with predominantly hemolytic PNH. LTC-IC numbers were extrapolated from secondary CFC numbers after 5 weeks of culture, and clonogenicity of LTC-ICs was determined by limiting dilution assays. When compared with normal volunteers (n = 13), PNH patients (n = 14) showed a 4.7-fold decrease in CD34⁺ cells and an 8.2-fold decrease in CFCs. LTC-ICs in BM and in PB were decreased 7.3-fold and 50-fold, respectively. Purified CD34⁺ cells from PNH patients had markedly lower clonogenicity in both primary colony cultures and in the LTC-IC assays. As expected, GPI-anchored proteins were decreased on PB cells of PNH patients. On average, 23% of monocytes were deficient in CD14, and 47% of granulocytes and 58% of platelets lacked CD16 and CD55, respectively. In PNH BM, 27% of CD34⁺ cells showed abnormal GPI-anchored protein expression when assessed by CD59 expression. To directly measure the colony-forming ability of GPI-anchored protein-deficient CD34⁺ cells, we separated CD34⁺ cells from PNH patients for the GPI⁺ and GPI⁻ phenotype; CD59 expression was chosen as a marker of the PNH phenotype based on high and homogeneous expression on fluorescent staining. CD34⁺CD59⁺ and CD34⁺CD59⁻ cells from PNH/AA patients showed similarly impaired primary and secondary clonogeneic efficiency. The progeny derived from CD34⁺CD59⁻ cells were both CD59⁻ and CD55⁻. A very small population of CD34⁺CD59⁻ cells was also detected in some normal volunteers; after sorting, these CD34⁺CD59⁻ cells formed normal numbers of colonies, but their progeny showed lower CD59 levels. Our results are consistent with the existence of PIG-A–deficient clones in some normal individuals. In PNH/AA, progenitor and stem cells are decreased in number and function, but the proliferation in vitro is affected similarly in GPI-protein–deficient clones and in phenotypically normal cells. As measured in the in vitro assays, expansion of PIG-A^– clones appears not be caused by an intrinsic growth advantage of cells with the PNH phenotype.