VERIFICATORY TECHNIQUES 221 



facts which they describe, respectively, are related in nature 

 in a specifiable way. 



But this does not meet the difficulty which is involved 

 in making predictions from hypotheses. Only the contentual 

 implications of the hypothesis count. But empirical gen- 

 eralizations are of no avail in making predictions from 

 hypotheses. For every hypothesis designates an entity 

 which is not certainly known to exist. Hence, in the very 

 nature of the case there can be no descriptive generalizations 

 expressive of the behavior of hypothetical entities. Every 

 hypothesis, considered as a postulate for the drawing of 

 implications, represents a contrary-to-fact condition. It 

 has been debated by philosophers whether any consequence 

 follows from a contrary-to-fact condition. Certainly no 

 strictly contentual implication can be drawn. It may be 

 insisted that an hypothesis designates not a contrary-to- 

 fact condition but merely a different-from-fact condition. 

 But in either case the empirical foundation for the gen- 

 eralization has disappeared, for the hypothesis refers to a 

 state of affairs which has never been observed and which 

 therefore has never been described in generalizations. 



It follows that there is no technique for prediction. Here, 

 again, resort must be had to more or less imperfect analogies. 

 If molecules have never been observed, one simply cannot 

 know how they will behave. But billiard balls in a bushel 

 basket can be observed and crude analogies can be drawn 

 from such behavior. Furthermore, operations of serial ex- 

 tension upon such objects, involving consideration of them 

 as getting smaller and smaller, and more and more elastic, 

 can be performed. To be sure, in this process of refinement 

 the objects will almost certainly lose many of their empirical 

 properties; but they will probably retain some of them, and 

 the nature of the operation of refinement should tell the 

 investigator roughly what these are. But every prediction 

 from the resultant entity can be no better than probable. 

 Necessary prediction from hypotheses seems impossible. All 

 that can be said is that predictions will be more or less 



