86 LEONELL C. STRONG 



The method employed for testing the data was by comparing 

 the probable errors of the massed observations of each tumor. 

 According to this method, any result that is three times its 

 probable error is usually considered significant. ^ In the present 

 experiment the results obtained were as follows: 



1. dBrB 786 negative : 80 reactions =±=5.74 or 9.23% ± 0.67 



2. dBrA 760 negative : 38 reactions ±4.05 or 4.76% ± 0.52 



Difference 4.47% ± 0.85 



The difference is thus 5.25 times its probable error. 



In the case of these two tumors, therefore, physiological 

 reactions are independent of histological characteristics. It 

 will be remembered (p. 77) that the two tumors appeared 

 histologically identical. 



b. Constant reaction potentiality. The objection may be 

 raised that perhaps the reaction difference is due to the fluc- 

 tuating activity of only one type of tumor cell — the dBrB type 

 being in one phase, the dBrA in another phase of a large rhyth- 

 mical process of change. To test out the disputed cyclic 

 changes of tumor activity, the whole data for the dBrB tumor 

 were grouped into three periods, making the number of obser- 

 vations in each class as nearly equal as possible without splitting 

 up the data from any one chart. The three groups represent 

 successive periods in the course of the experiment. 



The data bearing on supposed fluctuations of the tumor cell 

 are as follow^s: 



1st period 198 negative : 26 reactions ±3.23 or 11.61% ± 1.44 

 2nd period 309 negative : 32 reactions ±3.63 or 9.38%, ± 1-06 

 3rd period 282 negative : 22 reactions ±3.05 or 7.23%, ± 1.01 



Comparing the differences between the three periods we 

 obtain : 



Difference between: 



1st and 2nd 2.23% ± 1.79 or 1.24 times probable error 

 2nd and 3rd 2.15% ± 1.46 or 1.47 times probable error 

 1st and 3rd 4.78% ± 1.76 or 2.48 times probable error 



^ The formula used in the present experiment in computing the probable error 

 of a given ratio is the one proposed by Pearl in his paper on human sex ratios ('08). 



