The Method u 



way seemingly incomplete generalizations are 

 raised to a higher rank and given a validity 

 that a multitude of facts in one field would not 

 insure them. I do not mean to assert that 

 this method may not lead to error, but I do 

 claim that it is legitimate and will eliminate 

 the errors into which deductive reasoning 

 sinks. In every case where this method has 

 been used I have sought for parallel expres- 

 sions, and finding them, have assumed that I 

 was on the track of a general law which ought 

 to be formulated. The value of facts is the'reby 

 raised and laws are thrown into a more general 

 form than if a purely inductive method were 

 followed. Grant that the theory of parallelism 

 is correct, and the broader generalizations will 

 at least deserve consideration. They are not 

 the result of a guess, but of a method persisted 

 in through a long period. 



The following rules are applications of this 

 method and efficient checks to error: 



1. A principle cannot be accepted as the 

 basis of a working hypothesis until at least 

 one instance of its action is known. 



2. The r presumption is against an explana- 



