SCIENTIFIC DISCOVERY 189 



has had previous samples of it. By a sample of the unknown 

 is meant an idea, once conjectural and hypothetical in 

 character, but now verified to such an extent as to have been 

 incorporated into the known. Hence one's previous success- 

 ful anticipations become the foundations for new anticipa- 

 tions. But the samples, by themselves, cannot afford such a 

 basis, for one cannot presume that nature is everywhere 

 alike. The principle of the uniformity of nature, so often 

 considered to be the sole justification for inferences as to the 

 future, asserts not that nature is everywhere the same but 

 that there are pervasive and repeating structures. The realm 

 of events is much more constant in its structural features 

 than in its contentual features. Hence the principle of 

 analogy entitles one only to infer that there will be the same 

 type of connection between the present known and the 

 present unknown that there was between the past known 

 and the past unknown. It justifies only the assumption 

 that one will continue to meet with success if he attempts to 

 explain in terms of abstractions, serial operations, associa- 

 tions, causes and effects, microscopic elements, and so on — 

 but it does not tell him specifically what content these 

 explanatory notions will have. It is therefore very limited as 

 a principle of anticipation. It enables one to predict the 

 character of the unknown only upon the basis of these highly 

 general structural features of the known, and does not enable 

 him to anticipate anything as to its specific contentual fea- 

 tures. These hmitations justify the claim that there really 

 is no logic of discovery in the strict sense of the term. 



TECHNIQUES OF CONSTRUCTION 



The techniques of construction are not distinguished in any 

 significant way from the techniques of description, discussed 

 in Chapter VI. That the former are "elaborative" rather 

 than "descriptive" means only that they refer to the less 

 obvious features of events. Hence they are such operations 

 as are involved in classifying, ordering, and correlating 

 events when the properties determining these operations are 



