121 



the same cause on the volume of the smallest berry to 

 that on the largest one will be as 49 to 64. This diffé- 

 rence is still too small to cause any very marked skewness. 



Now such a smallness of the variations as compared 

 with the absolute size of the individuals, seems to be 

 rather the rule in nature. The conséquence will be, that, 

 though in reality the curves will be skew, the différence 

 from a normal curve will generally be very small. 



The same reasoning explains, how we very generally 

 fînd errors of observation distributed in normal curves. 

 For in nearly ail measurements the errors made will be 

 incomparably smaller than the quantity measured. 



There are some measurements however in which the 

 errors become quite of the order of the quantity sought. 

 Such for instance is the détermination by observation of 

 the threshold of sensation. Further on (see Example II, 

 art. 15, tab. 5, fig. 5) I will give a séries of measurements 

 of this quantity, which shows that just in this case 

 we fînd the errors of observation distributed according 

 to quite another law than that of the normal curve. 

 After what has been said, the fact has nothing very 

 surprising. 



11. Conclusion. Summing up, we fînd that causes inde- 

 pendent of the size of the individuals produce normal 

 curves, causes dépendent on this size produce skew curves. 

 The latter case must be the gênerai one. There seems 

 every reason to expect, however, that the skewness will 

 be exceedingly small in many cases. 



In several cases we feel at once that the effect of the 

 causes of déviation cannot be independent of the dimen- 

 sion of the quantities observed. In such cases we may 

 conclude at once that the frequency-curve will be a skew 

 one. To take a simple example: 



Suppose lOOOO men to begin trading, each with the 

 same capital; in order to see how their wealth will be 



