BCG originated from a virulent bovine strain of the tubercle bacillus after prolonged serial subculture on a potato medium. Since attenuation was achieved, the BCG strain has been distributed to a large number of centres where BCG vaccine is produced. Many of these production laboratories have maintained their BCG lines by continuing serial transfers, but have employed a variety of media for this purpose, and have produced BCG vaccine by a variety of techniques. Distinct differences have developed between some of the daughter strains of BCG, but the mechanism through which these changes have occurred has not been clear. In recent years methods have been developed which have enabled changes taking place within some BCG strains during experimental serial subculture to be monitored. In this survey the relationship of the changes observed to the different techniques employed for the maintenance of BCG lines and for the preparation of vaccine is considered. It is suggested that selection of minority populations within BCG strains noted during experimental studies may provide an analogy with the mechanism through which the original attenuation of the virulent bovine strain was brought about. The relevance of small-scale laboratory investigations to full-scale production procedures is also discussed, and finally some additional measures that might be taken to minimise changes in BCG strains are proposed.