ANALOGY. 629 



cubes ; therefore it is to be expected that potassium chlo- 

 ride will also crystallise in cubes. The science of chemistry 

 as now developed rests almost entirely upon a careful 

 and extensive comparison of the properties of substances, 

 bringing deep-lying analogies to light. When any new 

 substance is encountered, the chemist is guided in his 

 treatment of it by the analogies which it seems to present 

 with previously known substances. 



In this chapter I cannot hope to illustrate the all- 

 pervading influence of analogy in human thought and 

 science. All science, it has been said, at the outset, arises 

 from the discovery of identity, and analogy is but one 

 name by which we denote the deeper-lying cases of re- 

 semblance. I shall only try to point out at present how 

 analogy between apparently diverse classes of phenomena 

 often serves as a guide in discovery. We thus commonly 

 gain the first insight into the nature of an apparently 

 unique object, and thus, in the progress of a science, we 

 often discover that we are treating over again, in a new 

 form, phenomena which were well known to us in another 

 form. 



Analogy as a Guide in Discovery. 



There can be no doubt that discovery is most frequently 

 accomplished by following up hints received from analogy, 

 as Jeremy Bentham remarked. 1 Whenever a phenomenon 

 is perceived, the first impulse of the mind is to connect it 

 with the most nearly similar phenomenon. If we could 

 ever meet a thing wholly sui generis, presenting no 

 analogy to anything else, we should be incapable of 

 investigating its nature, except by purely haphazard 

 trial. The probability of success by such a process is 

 so slight, that it is preferable to follow up the faintest 

 clue. As I have pointed out already (p. 418), the pos- 

 sible experiments are almost infinite in number, and very 

 numerous also are the hypotheses upon which we may 

 proceed. Now it is self-evident that, however slightly 

 superior the probability of success by one course of proce- 

 dure may be over another, the most probable one should 

 always be adopted first. 



1 Essay on Logic, Works, voL viiL p. 276. 



