This review examines the evidence that bone marrow failure (BMF) in aplastic anaemia (AA) is due to loss of haematopoietic stem cells (HSCs), which, in turn, is caused by deranged immunity and inflammation. We also consider how the course of the disease and the response to immuno‐suppressive therapy are influenced by the nature and specificity of the pathogenic process. A somatic mutation of the PIGA gene underlies the clonal disease paroxysmal nocturnal haemoglobinuria (PNH): there is direct evidence that the expansion of the PIGA mutant clone results from Darwinian selection exerted by a glycosyl‐phosphatidyl‐inositol ‐specific auto‐immune attack. Thus, PNH patients are a unique subset of patients with AA, in whom haematopoiesis recovers through this escape mechanism. A similar process, although less effective, may operate when the auto‐immune attack is against a human leucocyte antigen (HLA) molecule and an HLA mutation has produced a clone missing that molecule. We then discuss the significance of other mutant clones that are frequently found in AA, presumably due to a combination of genetic drift and selection. These clones are not causative of AA, but they emerge in AA and they may be pre‐leukaemic: unlike a PIGA mutant clone, in general they are unable to effectively reconstitute haematopoiesis.