VARIABILITY 139 



knowledge of the variability of certain characters in certain races 

 has been obtained that is, a knowledge of the range within the 

 limits of which variations normally occur, the relative frequency of 

 each degree of deviation from the mode (the most common type), 

 and the degree in which the deviations of various characters tend 

 to be correlated with one another. Such knowledge is gathered 

 only at great expense of time and patience. It can, as a rule, 

 include only the variations of the fittest (the survivors), and never 

 more than a very few measurements in a very few animals and 

 plants. It may, of course, like knowledge of any other facts, prove 

 very useful. But curiously little effort has been made to use it, 

 to interpret it, to link it up with laws, with uniformities of causation. 

 Apparently the result, such as it is, of each inquiry has been held 

 to justify the labour. As far as I am able to judge, however, the 

 biometric evidence concerning variability is valuable chiefly for 

 the light it sheds on, and the support it affords to the hypothesis 

 that variations are both c spontaneous ' and under the control of 

 Natural Selection. It demonstrates that variations tend to occur 

 about equally about the specific mean, the plus and minus 

 variations of every dimension of a character being of nearly equal 

 frequency and magnitude ; material thus being afforded which 

 enables Natural Selection to meet every contingency. Again, 

 while all characters vary, large variations which tend to ruin the 

 co-adaptation of the parts of the organism are rarer than smaller 

 variations and most rare in the case of characters in which the 

 co-adaptation to the parts needs to^be very close ; thus the squirrel's 

 head and forefoot vary less widely in proportion as regards length 

 than the tail. Yet again, when characters are functionally corre- 

 lated, the variations tend to be correlated also ; thus if the length 

 of a bone in a man's arm is greater or less than the average, it is 

 more likely that the other bones in his arm will be correspondingly 

 long than that his teeth, nose, or beard will be so. If the Bathmic, 

 or the Lamarckian doctrine, or the hypotheses that variations are 

 commonly caused by the direct action of the germ's environment 

 were any of them true, variations would not be grouped about the 

 specific mean, but would all be in one direction or at least would 

 strongly preponderate in it. If variability were a * fundamental ' 

 property of living beings, uncontrollable by Natural Selection, all 

 characters should be equally variable. At any rate small variations 

 should not so considerably and universally preponderate in closely, 

 as compared to less closely, co-adapted parts. This last hypothesis, 

 however, is even more decisively negatived by the fact that the 



