4o6 THE GRAMMAR OF SCIENCE 



and that at the end of the time the types, variabihties, 

 and correlations of the same organs are again determined 

 for such of the individuals as still remain alive. The indi- 

 viduals will not be so numerous, and their reduction per 

 hundred of the original number is the total death-rate. 

 Further, the types, variabilities, and correlations, one, or 

 more, or all of these may or may not differ on the second 

 determination. If they differ, then the death-rate is said to 

 be partly selective ; if they do not differ, the death-rate is 

 non-selective. As a rule the death-rate is compound, and 

 in order to ascertain the effectiveness of selection we must 

 separate the total death-rate into selective and non-selective 

 portions. We require, then, to ascertain whether the 

 death-rate is a function of the quantitative measure of any 

 particular character. 



Now the task involved is a very difficult one, for 

 selection may have changed the type, the variability, or 

 the correlation of one or more organs. But we have seen 

 that characters and organs are more or less closely cor- 

 related among themselves. Hence in selecting one organ, 

 i.e. changing its type, death changes to a greater or lesser 

 extent all correlated organs. Again, if the type be left 

 constant, but the variability changed, this changes not only 

 the variability but also the correlation of all correlated 

 organs. Further, in some forms of frequency distribution 

 — skew distributions — it is not possible to change the type 

 or the variability or the correlation without changing the 

 other two quantities as well. An organism is in fact to 

 be looked upon as a whole, and it is impossible to change 

 one character without changing the whole system of cor- 

 related characters. If it were an advantage to man to 

 have a radius of a given length, and a death-rate produced 

 a race with a radius varying slightly about this length, 

 then we should find not only the radius but the humerus, 

 and not only the arm bones but the leg bones, the stature, 

 and even the rotundity of the skull changed. Further, 

 not only these organs would be changed in type, but in 

 variability and degree of correlation. It may be shown 

 that the correlation can even be changed in sign, and that 



