R. A. Fisher, H. G. Thornton, and W. A. Mackenzie :33o 



doing so, however, it was thought advisable to exclude as far as possible 

 the exceptional large values. As a rough criterion it was decided to 

 exclude those values which exceeded by more than threefold the mean 

 value of the group. In the larger groups this criterion acted well; in the 

 smaller groups, such as occurred for high and low values of .'•, it was 

 necessarily inconclusive, even when account was taken of neighbouring 

 groups. The curve fitting was therefore confined to the region in which 

 the data appeared to be sufficiently abundant. 



30 



60 



90 



120 



150 



180 



Fie-. 1. Kiiiootli curves fitted to Cutler's data. 



Curves of the form v = Ax + Bx^ 



(where A and B are two constants determined from the data) were fitted 

 to the four-plate data from ,f = to ir = 180, and to the five-plate data 

 from to 160; the curves obtained are shown in Fig. 1. 



The straight fine, v = x, represents the relation between the variance 

 and the mean in the Poisson Series. The curves evidently tend to cling 

 closely to this line, especially in the region (60-120) where the data are 

 most abundant. The curves strongly suggested that the departures in 



