38 



Art. 9.— T. Terada 



dispersed rather irregularly as shown in Fig. 11 in which 

 the results are grouped with respect to the hours of occurrence. 

 Next, the results were grouped according to the periods and the 

 mean ratios were calculated for each group of the periods. The 

 results are shown in the following Tal:)le and also plotted in 

 Fig. 12 where tlie mean periods^^ are taken as abscissa. 



Fig. 12. 



AX 



0% 



0-6 



04 



Table IV. 



As will be seen, the ratio increases at first rapidly with the 

 periods, then gradually tends to an asymptotic value probably 

 nearly equal to unity. The fact is rather remarkable and must 

 serve as an important basis for explaining the phenomena in 



1) The mean of the actual periods corresponding to the trains taken was calculated, not 

 the simple mean of the interval of the period concerned. 



