264 THE EVOLUTION OF SCIENTIFIC THEORIES 



as well or better, we will stubbornly defend the threatened theory. 

 Having obsers^ed the inexhaustible resources that can be brought to 

 the defense of any given theory, we see that no crucial experiment 

 can force on us a final absolute falsification of that theory. But what 

 such an experiment may well do is to sway our judgment to a verdict 

 that, compared with another, the given theory is relatively less 

 sufficient. That is, ultimately we accept falsification of die theory 

 when we consider it has been shown functionally inferior to a com- 

 petitor. 



Three Functional Criteria 



All possibility of strictly logical verification or falsification being 

 denied us, we must in the end look to the functional criteria. Com- 

 plex, uncertain, in a degree irreducibly subjective, they are at least 

 usable criteria. 



CORRELATIVE EFFICIENCY 



Correlative efficiency seems the most straightforward of the criteria 

 by which we might appraise a scientific theory. We might for exam- 

 ple conceive a correlative index thus : 



Number of 

 independent colligative relations 



accommodated in theory „ 

 = ^=7 



Number of 



independent postulates required 



to constitute the theory 



Can we not always choose with confidence, among competing the- 

 ories, that one with the highest value of the index ratio? But which 

 does offer the better index ratio: Avogadro's theory, which uses more 

 postulates to correlate more relations; or Berzelius' theory, which 

 uses fewer (and less extreme) postulates to correlate fewer rela- 

 tions? One might then wish for quite precise evaluations of the index 

 ratios, and I do not believe any such evaluations possible. 



Can we even estimate the numerator for any major theory? And 

 were we given some such estimate of number, could it mean any- 

 thing if we did not attach to each relation some factor indicative of 



