THE LAW OF REFRACTION 407 



and on reference to the tables P = -53, i.e. in 53 out of 100 cases 

 we should get in random samphng a fit as bad or worse. Or, 

 in other words, the theoretical formula fits the observations 

 quite well. We should not, however, expect a difference between 

 theory and experiment to show itself, unless the observations 

 were made with the utmost refinement possible. 



Truth in physics is only a measure of probability. This 

 we have known for long. But it is only now that methods are 

 opening up by which we are able to estimate numerically the 

 truth of physical laws, methods which will doubtless afford 

 profitable research to many workers. 



