260 THE EVOLUTION OF SCIENTIFIC THEORIES 



the syntactic rules been properly applied? Does the derivation in- 

 volve approximations or assumptions, failing in the (nonideal) situa- 

 tion 11, without which we actually attain the relation C? Again we 

 have manifold opportunities for remedial action, but again those 

 opportunities are bounded: we can permit ourselves no action that 

 impairs the theory's capacity to construe the many other relations 

 it has earlier handled in convincing fashion. 



And so at last we may be driven to critical re-examination of the 

 theory itself. Ordinarily we will find it difficult to alter the funda- 

 mental axioms, to deal with situation 11, without producing new 

 complications elsewhere in the theoretical configuration. However 

 we can always save the theory by odding to its postulates one or 

 more ad hoc assumptions expressly designed to substitute Cn for 

 Cii. With such additions we construct a new theory (AfF)' which 

 contains the complete essence of the old. Thus the core of the Ptole- 

 maic system— a stationary earth and celestial motions compounded 

 from circles— was maintained by the addition of further epicycles, 

 equants, and eccentrics. Nor is it only fundamentally unsound the- 

 ories that require such salvage operations. Note diat Copernicus' 

 own system deploys some thirty-odd epicycles. For benzene Kekule 

 is forced to postulate bonds with a wholly unprecedented degree of 

 lability, in order thereby to reconcile his valence theory with the 

 observed isomerism of substituted benzenes. In our own era the 

 neutrino is called into being to maintain the conservation of energy 

 in beta decav. 



Gi\^en such distinguished precedents, we may then well elect to 

 bolster a prepossessing theory with one or several "saving graces." 

 This possibility is always open to us; and beyond it looms yet an- 

 other, entirely new, line of defense. Without making any alteration 

 whatever in the theory we can perfectly reconcile it with the chal- 

 lenging fact! For of course the ramified chain of reasoning which 

 links fact with theory involves not just one theory but, as Weyl notes, 

 a whole galaxy of other (potentially alterable) theories and relations. 



Individual scientific statements cannot be ascribed an intuitively veri- 

 fiable meaning, but truth forms a system that can be tested only in its 

 entirety. 



The totality of what is tested. Pascal considered the Puy de Dome 

 experiment a decisive disproof of the theory that Nature abhors a 



