of mean or median survival time. of time. In such cases the effect is given in units Alternatively, the effect could be given as a percentage of control values and in such cases the percent shortening of the mean or median age at death or the percent shortening of the mean or median survival time would be the relevant parameters. It should be noted that the percentage effect as measured by the shortening of the mean or median age at death is not equivalent in most cases to that measured as precent shortening of the mean or median survival. Although it would have been desirable to use the same way of expressing the effect throughout the present document, it was impossible to do so owing to the lack of suitable data in the documents reviewed. 14, Mean and median life-span are the average duration of life experienced by the animal population and the time required for 50 per cent of the animals to die, respectively. Thez do not offer any indication of the variability affecting this phenomenon in time: they are therefore, as such, unsatisfactory parameters for any statistical analysis. The curve describing the extinction of the population in time is more informative in that it allows knowledge of the time when this process begins and ends and it also allows one to know whether it has taken place regularly. Irregularities of the curve may some- times be attributed to specific causes or set of causes. Both the mean and the median life-span and the curve of per cent cumulative mortality can be readily calculated. 15. The age-specific mortality rate is a more elaborate parameter: it ex- presses the instantaneous rate of mortality of the animals at risk as a function of age. Changing of this parameter in time is therefore its main dis- advantage: the advantage lies in its sensitivity in measuring the changes in the distribution of times at death. It should be recalled that the displace- ment of the age-specific mortalityrate curve for irradiated animals above the curve for non-irradiated controls does not reflect days of life lost, but the increased rate of dying at a given age. time and any irregularity in it are extremely The trend of this parameter in useful to identify possible specific causes of death. 16. In estimating mortality rates it is desirable that assessments be inde- pendent of the proportion of animals that have died by any given time, i.e. that the estimate should be truly non-parametric. To this end, different