THE DEDUCTIVE METHOD. 515 



warrant reliance on the general conclusions arrived at by 

 deduction, these conclusions must be found, on careful com- 

 parison, to accord with the results of direct observation 

 wherever it can be had. If, when we have experience to com- 

 pare with them, this experience confirms them, we may safely 

 trust to them in other cases of which our specific experience 

 is yet to come. But if our deductions have led to the conclu- 

 sion that from a particular combination of causes a given effect 

 would result, then in all known cases where that combination 

 can be shown to have existed, and where the effect has not 

 followed, we must be able to show (or at least to make a pro- 

 bable surmise) what frustrated it : if we cannot, the theory is 

 imperfect, and not yet to be relied upon. Nor is the verifi- 

 cation complete, unless some of the cases in which the theory 

 is borne out by the observed result, are of at least equal com- 

 plexity with any other cases in which its application could be 

 called for. 



If direct observation and collation of instances have fur- 

 nished us with any empirical laws of the effect (whether true 

 in all observed cases, or only true for the most part), the most 

 effectual verification of which the theory could be susceptible 

 would be, that it led deductively to those empirical laws ; 

 that the uniformities, whether complete or incomplete, which 

 were observed to exist among the phenomena, were accounted 

 for by the laws of the causes <were such as could not but exist 

 if those be really the causes by which the phenomena are pro- 

 duced. Thus it was very reasonably deemed an essentia 

 requisite of any true theory of the causes of the celestial 

 motions, that it should lead by deduction to Kepler's laws : 

 which, accordingly, the Newtonian theory did. 



In order, therefore, to facilitate the verification of theories 

 obtained by deduction, it is important that as many as pos- 

 sible of the empirical laws of the phenomena should be as- 

 certained, by a comparison of instances, conformably to the 

 Method of Agreement: as well as (it must be added) that 

 the phenomena themselves should be described, in the most 

 comprehensive as well as accurate manner possible; by col- 

 lecting from the observation of parts, the simplest possible 



332 



