THE DISCOVERY OF LAWS 67 



necessarily from the subject-matter of science and to 

 which, therefore, a law can have no reference. It is 

 quite possible that, if such a case as we are imagining 

 actually occurred, we should adopt this course ; but 

 it must be remembered that we might adopt the other 

 and remove the discrepancy, not by rejecting the obser- 

 vation, but by stating two laws. Which alternative 

 we shall adopt depends on all the circumstances, and 

 here it is convenient to note why the observation of a 

 very large number of favourable instances is important 

 in the establishment of a law. If we have based a law 

 on a large number of instances, and subsequently find 

 other instances apparently discrepant, then, if, when 

 we choose between the alternatives just mentioned, we 

 reject the law, we place all these large number of obsep 

 vations outside the province of science. And this wtT* 

 are loath to do ; we want to reduce as much as possible 

 of our experience to order by means of laws, and the 

 rejection of the whole of our past experience as one 

 great " mistake," accords ill with that purpose. When 

 we have ordered a very large number of instances by 

 means of a law, we shall want to maintain that law at 

 all hazards ; and we shall be much more willing to 

 introduce other laws to Jnclude instances apparently 

 discrepant, and so to avoid the necessity of rejecting 

 tEe material on which the original law was based, than 



O 



we should be if we have only ordered a very small number 

 of instances. 



THE PURPOSE OF LAWS 



It will be seen that in this discussion the question from 

 which it started has almost been left out of account. 

 We asked how we managed to establish laws, by the 

 examination of our past experience, which were true 

 also for future experience. The considerations that 

 have been put forward suggest that this problem is not 



