Paroxysmal nocturnal hemoglobinuria (PNH) is characterized by the expansion of one or more clones of stem cells producing progeny of mature blood cells deficient in the plasma membrane expression of all glycosyl phosphatidylinositol (GPI)-anchored proteins (AP). This is due to somatic mutations in the X-linked gene PIGA, encoding one of the several enzymes required for GPI anchor biosynthesis. More than 20 GPI-APs are variously expressed on hematological cells. GPI-APs may function as enzymes, receptors, complement regulatory proteins or adhesion molecules; they are often involved in signal transduction. The absence of GPI-APs may well explain the main clinical findings of PNH, i.e., hemolysis and thrombosis in the venous system. Other aspects of PNH pathophysiology such as various degrees of bone marrow failure and the dominance of the PNH clone may also be linked to the biology and function of GPI-APs. Results of in vitro and in vivo experiments on embryoid bodies and mice chimeric for nonfunctional Piga have recently demonstrated that Piga inactivation confers no intrinsic advantage to the affected hematopoietic clone under physiological conditions; thus additional factors are required to allow for the expansion of the mutated cells. A close association between PNH and aplastic anemia suggests that immune system mediated bone marrow failure creates and maintains the conditions for the expansion of GPI-AP deficient cells. In this scenario, a PIGA mutation would render GPI-AP deficient cells resistant to the cytotoxic autoimmune attack, enabling them to emerge. Even though the 'survival advantage' hypothesis may explain all the various aspects of this intriguing disease, a formal proof of this theory is still lacking.