118 Proceedings of the Royal Society 
duration of the risks, showing that the light mortality during the 
years while selection is in operation is balanced by a heavier mor¬ 
tality thereafter, and showing further that that heavier mortality 
is considerably greater than the general average mortality of a 
single aggregate mortality table. These are exemplified in various 
forms. The comparisons, however, are all based upon lives once 
assured. There is, finally, given in one view the rate of mortality 
experienced on all entrants of each age, during each year of assur¬ 
ance, as the true exponent of the rate of mortality on assured lives, 
along with five abridgments of the same, in the case of persons 
assuring at each quinquennial age. 
Causes of Death. 
There is also given the intensity of the causes of death at each 
age, and the relation of the deaths of assured lives, from various 
causes, to the deaths of the male population of Scotland, pointing 
out the several orders of disease in which the mortality of assured 
lives is greater or less than the population. There is also given, in 
a general form, the effects of selection upon the various causes of 
death, pointing out those in which selection appears to have been 
of greatest benefit. 
Diseased Lives. 
The usual comparisons are made of the actual with the computed 
number of deaths, and also with the rates of mortality, pointing out 
that the mortality on diseased lives is greater than on healthy lives 
by about 20 per cent. The diseased lives were thereafter broken 
up into sections, according to the nature of the imperfections for 
which the extra charge was made, and showing the rate of mortality 
experienced on four such classes. For two of these classes—un¬ 
favourable personal history and gout—and also for the general class 
of diseased lives, the law of mortality is given, as well as the 
annual premium for assurance of L.100 at death, showing the extra 
charge for such classes of lives. 
Years of Life. 
Ail the foregoing methods of comparing actual with computed 
results have dealt with numbers of deaths. A method is pointed out 
