I find that writing about the safety of nuclear reactors becomes more difficult with t ime-no t because reactors are more dangerous than 5-10 years ago; but, because much more is known about reactors, propositions regarding their safety become more detailed and sophisticated. At one time it was said, inter alia, that magnox reactors were safe because they could tolerate complete loss of coolant flow and thereafter the fuel would radiate its fission product decay heat to the graphite moderator . I t was said of water reactors that a complete severance of the largest pipe in the pr imary circuit could be accepted by a strong containment and the system would remain safe. These points were important at the time but were only highlights which, taken out of context, could give false reassurance. It is as though safety were claimed for a two-engined aircraft on the grounds that it could fly on one engine--i .e, a useful attribute but a somewhat superficial observation. Even in recent times I have described some of the safety promoting features of a sodium-cooled fast reactor as; (Farmer, 1971b, c; Farmer and Gilby, 1967; UKAEA Rep. AHSB(S)(R), 1970): its use of a low pressure coolant having a very high thermal conductivity its stability associated with a negative power coefficient its capacity to store heat in the coolant equivalent to many hours of decay heat its capabil i ty of rejecting decay heat to the atmosphere with little or no requirement for external power or water. These are useful features which "can be used to promote safety and if used wisely could lead to the fast reactor becoming the safest of all systems" (U.K.4.E.A. Rep. AHSB(S)(R), 1970). However, I recognize that these features taken in isolation do not ensure the safety of a part icular design any more than in a reverse situation a reference to a nuclear excursion denies the possibility that fast reactors can be safe. The difficulty in writing about reactor safety stems from the fact that the safety of a nuclear reactor (or, for that matter, any sophisticated modern engineering complex) requires a thorough appreciation of the behaviour of the reactor; requires good design, inspection etc., and consequently any argument which sets out the basis for the safety of the plant must necessarily be detailed and lengthy. A safety report for a power reactor will give a simple description of the reactor and will indicate its response to the varying demands which arise during normal operation, and will also consider many abnormal conditions and give a fair description of their consequences on the reactor. The abnormal conditions may be faults arising through failure of equipment, faults imposed through maintenance or during operation, or imposed by external agencies-flood, fire, earthquake etc. The list of faults which are considered tends to increase year by year, partly through a greater awareness of possible fault, and partly through the assumption of more unusual conditions or those having a low probabil i ty of occurrence, as, for example, high floods, tornadoes etc. Not only do we have more fault conditions to examine, but there is a general demand for more convincing evidence that the reactor will safely survive these events, even those of low probabili ty. The use of the word "safely" introduced a further difficulty--how to define or reach agreement on "How Safe is Safe ?" We are surrounded by potentially hazardous situations (USAEC Rep. WASH1250). There is risk in travelling, eating, drinking, building houses, roads, factories, and risk in making or using both naturally occurring and man-made products, including power. Any activity carries with it a degree of risk and, for example, tbere is a risk of: Death from traffic accidents Death from drowning more...