210 THE VARIATION OF ANIMALS IN NATURE 



would hesitate to allow that Bumpus's case was proved, the 

 action of selection is likely.' He stresses the fact that the 

 number of individual variates is low, and is clearly divided 

 between an adherence to a rigid statistical principle (which, 

 when applied to the data, gives ' significant ' differences in 

 some characters) and an apprehension that, on account of 

 the paucity of data, the statistical principle may be fallacious. 

 Incidentally we may note that we have applied the current 

 tests to Bumpus's figures as a check on Harris's procedure 

 and find that his conclusions as to ' significance ' are 

 valid. 



The matter might be left to remain in this rather unsatis- 

 factory condition, with the admission that, statistically at 

 least, Bumpus's conclusions are sound. But there is, however, 

 a further question to be decided, which we think invalidates 

 these observations at their source. As one of us has pointed 

 out (Robson, I.e. p. 214), the cause of the death of the elimi- 

 nated is uncertain. What Bumpus did was to compare the 

 birds which recovered with those which died after being 

 blown down. All the birds were, it is admitted, blown down 

 by the gale ; but those which did not recover might have 

 died from various causes {e.g. from dashing in their fall against 

 a stone or a tree, from exposure and starvation, from the 

 immediate effects of strain and exhaustion). In short, the 

 birds might, we agree, be all blown down on account of some 

 structural deficiency, but their survival or death after failure 

 to sustain themselves in the gale might very easily be determined 

 by quite a distinct set of causes. In short, we are plainly 

 dealing with two distinct phenomena- — the fact of being blown 

 down on the one hand, and the multiple causes of death 

 connected with the subsequent experience of those who were 

 blown down. It might be urged that the acid test is really 

 between death and survival — that at all events we know there 

 were some significant differences between those which died and 

 those which survived. But in reply we must, obviously, ask 

 how any structural character (such as weight, wing spread, 

 etc.) which might determine whether a bird was blown down 

 or not, could determine whether a bird survived or died after 

 it was blown down — a result which might be determined by such 

 purely accidental causes as whether it hit a branch or stone in 

 its fall, or whether it was able to withstand exposure and shock. 



