370 Dr. Dorothy Wrinch and Dr. H. Jeffreys on Certain 



regarded the scientific practice of proceeding from the parti- 

 cular to the general as formally fallacious : as indeed it is, 

 if no postulate be introduced other than those of pure logic 

 and the bare facts of sensation. Scientific writers, on the 

 other hand, start with a firm conviction that their methods 

 are valid ; accordingly, if philosophical argument is opposed 

 to science, they regard the fact merely as a good, reason 

 for condemning philosophy. The result has been that 

 they look with disfavour on any analysis of the funda- 

 mental assumptions of science, since such discussions have 

 proved almost entirely fruitless in the past. Thus any 

 serious attempt at such analysis has come to be described 

 as metaphysical and largely ignored, with a consequent 

 loss of clarity in scientific discussion. A certain amount 

 of attention has, however, at last been drawn to this need, 

 partly by the slashing attacks of Karl Pearson and Ernst 

 Mach on certain prevalent scientific concepts, and partly 

 by the important physical results predicted by Einstein, 

 largely based on the views of these earlier writers. 

 Physical concepts have consequently been subjeeted to 

 some discussion in recent books, but, we think, quite 

 unsatisfactorily. Instead of investigating the actual nature 

 and method of application of the fundamental postulates, 

 these writers have surrendered completely to the philosophic 

 criticisms and tried to treat scientific knowledge without 

 using any form of generalization, but with the introduction 

 of certain new postulates which, if they occur in ordinary 

 scientific use at all, are not primitive propositions. The 

 result is that beautifully coherent deductive systems are 

 obtained which would be perfectly satisfactory if their 

 fundamental postulates were admitted ; but when these 

 are examined it is found in all cases that they are not 

 directly known to be true, and that they can be verified 

 only by the confirmation of the predictions based on them. 

 But this does not prove that they are true : if two pro- 

 positions p and q are so related that p implies q, the fact 

 that q is true does not entitle us to say that p is true, 

 unless some further assumption is introduced; and this 

 assumption is not a part of pure logic. Thus the truth 

 of the alleged primitive postulates in these cases is only 

 inferred by using a principle which the purely logical 

 method was expressly designed to avoid. 



One type of postulate that is unsatisfactory on these 

 grounds is that which involves the use of infinite classes 

 of entities, and our reason for objecting to it is similar to 

 that given by Poincare in his criticisms of Cantor's theory 



