2 DARBISHIRE, Tables illustrating Statistical Correlation 



that so long as we are as much in the dark as we are at 

 present about the circumstances which may affect an 

 animal's or plant's chances of attaining maturity, any 

 statement that such and such a feature matters or does 

 not matter is unwarrantable. 



But let us return to our argument. Some living things 

 are variable. We may adopt two attitudes, towards this 

 variability ; we may either say that it does not matter 

 and ignore it, or we may suspect that it may matter and 

 measure it. In my opinion evidence does not justify 

 us in adopting the former attitude. Statisticians have 

 provided us with a method for measuring this variability. 

 But we usually want to know more than this ; we want, 

 if possible, to measure the closeness of the relation 

 between two such variable things. Statisticians have 

 again provided us with a method which enables us to 

 measure the closeness of that relation in which biologists 

 are most interested, namely, that between parents and 

 children. 



The first step in this method is to construct a 

 Correlation Table. How this is done is best explained 

 by giving an account of Weldon's beautiful experiment. 

 The variable phenomenon he dealt with was the number 

 of dice, in a throw of 12, which fell so that faces with 4 or 

 more pips on them were uppermost. When we throw a 

 single die it is an even chance whether it falls so that a 

 face with 3 or fewer pips on it lies uppermost, or whether 

 a 4-or-more-bearing face lies uppermost. Therefore the 

 most probable number of dice with faces bearing 4 or 

 more uppermost in a throw of 12 is 6, but the number 

 may be anything between o and 12 inclusive, though 

 these extreme results occur very seldom. Here is a list 

 shewing the frequency with which the 13 possibilities 

 occurred in a thousand throws which I have made : 



