Can the carcinogenic risks of radiation that are observed at high doses be extrapolated to low doses? This question has been debated through the whole professional life of the author--now nearing four decades. In its extreme form the question relates to a particular hypothesis (LNT) used widely by the international community for radiological protection applications. The linear no-threshold (LNT) hypothesis propounds that the extrapolation is linear and that it extends down to zero dose. The debate on the validity of LNT has increased dramatically in recent years. This is in no small part due to concern that exaggerated risks at low doses leads to undue amounts of societal resources being used to reduce man-made human exposure and because of the related growing public aversion to diagnostic and therapeutic medical exposures. The debate appears to be entering a new phase. There is a growing realisation of the limitations of fundamental data and the scientific approach to address this question at low doses. There also appears to be an increasing awareness that the assumptions necessary for a workable and acceptable system of radiological protection at low doses must necessarily be based on considerable pragmatism. Recent developments are reviewed and a historical perspective is given on the general nature of controversies in radiation protection over the years. All the protagonists in the debate will at the end of the day probably be able to claim that they were right!