The scientific framework that has evolved for the protection of humans from ionizing radiation is based on a number of related features, including the use of reference anatomical and physiological models for the assessment of radiation doses from external and internal sources; studies of radiation effects at the molecular and cellular level; a large range of experimental animal studies; plus epidemiological studies of exposed populations over many decades. Models and data have been used to derive tabulated, standardised data on the com­mitted ‘‘dose per unit intake’’ of different radionuclides for internal exposures, and ‘‘dose per unit air kerma or fluence’’ for external exposures of workers, patients and the public. Epidemiological and experimental studies have been used in the estimation of risks associated with external and internal radiation exposure. For biological effects, the data from human experience have been further supported by experimental biology. For cancer, and for heritable effects, the starting points are the results of epidemiological studies and of studies on animal and human genetics. These results are, in turn, supplemented by information from experimental studies on the mechanisms of carcinogenesis and heredity, in order to provide risk estimates at the low doses of interest in radiological protection.

In interpreting these data, certain balances have to be struck. With regard to radiation weighting factors, those for photons, electrons and muons are assigned a radiation weighting factor of 1. This is a simplification, particularly for photons, but is considered sufficient for their use in equivalent and effective dose terms because these are used for dose limitation, and assessment and control, in the low dose range. With regard to protons, external radiation sources are of most concern, and a radiation factor of 2 is used. A factor of 2 is also used for pions. These are particles of importance for exposures in aircraft, and for those involved with high-energy particle accelerators. Alpha particles are particularly important with regard to internal exposures, and a weighting value of 20 is used. A value of 20 is also used for fission fragments, which are also of importance with regard to internal exposures, and the same value is used for heavy ions, which are encountered in high altitude aviation and space exploration. Finally, neutrons are treated somewhat differently, and the radiation weighting factor for them differs in relation to energy over a range of about 2.5 to slightly over 20.

Similarly, balances have to be struck in view of the uncertainties surrounding the values of tissue weighting factors and the estimate of detriment. Thus it is currently considered appropriate, again for radiological protection purposes, to use age and sex averaged tissue weighting factors and numerical risk estimates. For stochastic effects, after exposure to radiation at low dose rates, nominal probability coefficients for detriment-adjusted cancer risk of 5.5 x 10 2 Sv [39] for the whole population and 4.1 x 10 2 Sv 1 for adult workers have been derived. For heritable effects, the detriment-adjusted nominal risk in the whole popu­lation is estimated at 0.2 x 10 2 Sv 1 and in adult workers at 0.1 x 10 2 Sv 1.

These risk estimates are called ‘‘nominal’’ by the ICRP because they relate to the exposure of a nominal population of males and females, with a typical age distribution, and are computed by averaging over age groups and both sexes. The dosimetric quantity recommended for radiological protection, the effective dose, is also computed by age and sex averaging. There are many uncertainties inherent in the definition of nominal factors to assess effective dose, but the estimates of fatality and detriment coefficients are considered adequate for radiological protection purposes. Nevertheless, as with all estimates derived from epidemiology, the nominal risk coefficients do not of course apply to specific individuals. For the estimation of the likely consequences of an expo­sure of a known individual, or of a known population, it is necessary to use specific data relating to that exposed individual or population.

In those situations in which the dose thresholds (100 mSv in a year) for deterministic effects in relevant organs could be exceeded, protective actions should be taken. At radiation doses below around 100 mSv in a year, the increase in the incidence of stochastic effects is assumed to occur with a small probability, and in proportion to the increase in radiation dose over the background dose.

In terms of managing exposure to radiation, it is also necessary to consider that, on the one hand, individuals may be simultaneously exposed to several sources, and thus an ‘‘individual-related’’ assessment of the total exposure has to be attempted; whereas, on the other hand, it is also necessary to consider all of the individuals exposed by a single radiation source or group of sources — a ‘‘source-related’’ assessment. Of the two, the primary importance, of course, is the source-related assessment, because action can be taken for a source to assure the protection of all individuals from that source.

In terms of presenting the scientific framework of radiological protection to a wider audience, the probabilistic nature of stochastic effects and the properties of the LNT model make it impossible to derive a clear distinction between ‘‘safe’’ and ‘‘dangerous’’. This inevitably creates some difficulties in explaining the control of radiation risks. The major policy implication of the LNT model is that some finite risk, however small, must be assumed, and a level of pro­tection therefore has to be established that is based upon what is deemed acceptable by society at any one time. This is the problem that the ICRP’s system of protection attempts to address.