292 Group Discussion 



obtain all the time, regardless of the cause of death. If you are con- 

 cerned with a standard value or reference against which you can com- 

 pare, why not use an animal, as you do in the laboratory, from 

 which you can get such data? I have been amazed that for over ten 

 generations in the housefly, for example, one can continue to get the 

 same mean value. I say this value is a good basis for comparison in 

 an experimental study. With the white rat, under specific laboratory 

 conditions, for example, the only thing that will vary in irradiation 

 tests will be the extent or dosage of radiation. With humans we run 

 into another problem because we always deal with life expectation 

 based on a population that is not always an identical cohort, but for 

 experimental studies we can get cohorts of an identical nature for 

 inbred lines of a number of species of lower animals. I do not see 

 why the mean longevity is not as good a criterion as anything else ; 

 it considers the accidental deaths, it considers the possible disturb- 

 ance that the mode would have from having had early deaths or 

 accidental deaths, and even a tail at the end resulting from the 

 extended longevity of the few long-lived cohort members. 



Perks: I am sure you are right for your problem, but diff'erent 

 problems require different solutions. 



Rotblat: If we say, for example, that radiation causes a contraction 

 of the scale of life, and suppose we are dealing with the ideal case in 

 which all animals die of old age, then for the irradiated animals we 

 would obtain the same curve but bodily shifted to the left. This 

 would be very nice, but in practice it may not be so. It may be that 

 radiation has caused earlier deaths without changing the end-point. 

 The curve would then change completely. Which of these will 

 actually happen depends upon the effect that radiations have on the 

 lifespan. 



Sacher: The average is certainly the first quantity to use in the 

 experimental situation, but you cannot characterize all the effects 

 of radiations, or of any other environmental influence, in terms of a 

 single parameter. Empirically you can then proceed to the succeed- 

 ing central moments. The question is to find out what parameters 

 of the survival curve are being influenced by the particular environ- 

 mental factors under investigation. I have pointed out how, in 

 studies of radiation effects on mice, you could characterize the eff'ects 

 of radiations in terms of the A and the a parameter of the Gompertz 

 equation [q^ = Ae°^]. A single dose of radiation — to restate what I 

 said yesterday — causes a change in A, without a change in a. Con- 

 tinuous exposure causes a change in the a coefficient. It is perfectly 

 true that this Gompertz equation is not an entirely adequate des- 

 cription of the life-tables of natural populations, but it should be 



