NATURAL SELECTION 195 



process acting purely on the adult can have little effect. It 

 is true that selection amongst larvae (so far as this heavy 

 death-rate is not purely random) will tend to produce unex- 

 pected results in the adult stage, the most numerous types of 

 the latter being chosen for the characters they bore as larvae 

 and not for their actual facies. But this will not avert the 

 effect of selection amongst the adults (see Fisher, 1930, p. 134). 

 If there is a differential death-rate amongst the adults, a certain 

 genotype will be favoured, and this form will occur in an 

 increased proportion amongst the larvae. As long as the 

 incidence of larval mortality does not actually tell against the 

 adult character, then, on the theory of chances, the survivors 

 of the larval holocaust will still show on the average the same 

 increased proportion of the adult genotype. 



The real conclusions that should be drawn from such 

 studies as those we have mentioned appear to be the following : 



a. Most animals — all those with a high rate of repro- 

 duction — have a very high mortality, especially in 

 the early stages. 



/?. This mortality often appears to be random : but the 

 appearance may be deceptive, and certainly a random 

 death-rate cannot as yet be directly verified. 



y. However large the random death-rate may be, it cannot 

 nullify the effect of any selective death-rate, even if 

 very much smaller. This is at least true when two 

 populations in competition are both of considerable 

 size, and is necessarily a result of the random nature of 

 the main death-rate — i.e. the proportions of each 

 form can be influenced only by death-rates which 

 are not random. Actually, if one population were 

 very small, as when a rare mutant competes with the 

 dominant type of a species, a large number of trials 

 might be necessary before the inherent impartiality 

 of the random process was actually observed — i.e. the 

 mutation might have to occur often enough for the 

 mutant individuals in the aggregate to form a fairly 

 large population. 



8. The only satisfactory way to investigate whether death- 

 rates are selective or not is to study in nature the 

 actual death-rates of competing forms, whether 

 species, varieties or mutants. 



