2 



SCIENCE. 



[N. S. Vol. III. No. 53. 



its examination, he searches for these par- 

 ticular features. If they are found the the- 

 ory is supported; and in case the features 

 tlius predicted and discovered are numerous 

 and varied, the theory is accepted as satis- 

 factorJ^ But if the reexamination reveals 

 features inconsistent with the tentative the- 

 ory, the theory is thereby discredited, and 

 the investigator proceeds to frame and test 

 a new one. Thus, by a series of trials, in- 

 adequate explanations are one by one set 

 aside, and eventually an explanation is dis- 

 covered which satisfies all requirements. 



When the subject of study is one of wide 

 interest it usuallj' happens that several in- 

 vestigators cooperate in the invention and 

 testing of hypotheses. Often each investi- 

 gator will originate a hypothesis, and a 

 series of rigorous tests will be applied 

 through the endeavor of each one to estab- 

 lish his own by overthrowing all others. 

 The different theories are rivals competing 

 for ascendancj', and their authors are also 

 rivals, ambitious for the credit of discovery. 

 The personal factor thus introduced tends 

 to bias the judgment and is to that extent 

 unfavorable to the progress of science ; but 

 the conflict of theories, leading, as it event- 

 ually must, to the survival of the fittest, 

 is advantageous. Fortunately there is a 

 mode of using hypotheses which regulates 

 the personal factor without restricting the 

 competition of theories, and this has found 

 favor with the greatest investigators. It 

 has recently been formulated and ablj^ ad- 

 vocated bj' our fellow-member. Prof T. C. 

 Chamberlin, who calls it the ' method of 

 multiple hypotheses. '* 



In the application of this method the stu- 

 dent of a group of phenomena, instead of 

 inventing and testing hypotheses one at a 

 time, devises at an earlj^ stage as many as 

 possible, and then, treating them as rival 

 claimants, assigns to himself the role of 



*The Method of Miiltiple Working Hypotheses, 

 Science (1st series), Vol. XV. (1890) pp. 92-96. 



judge. Returning to the study of nature, 

 he seeks for special features which cannot 

 consist with all the hypotheses, and may 

 therefore serve to discriminate among them. 

 Thus by a series of crucial tests he elimi- 

 nates one after another of the tentative the- 

 ories until but a single one remains, and he 

 then proceeds to apply such tests as he may 

 to the survivor. 



In these methods of work, w^hether theo- 

 ries are examined successively or simulta- 

 neously, there are two steps involving the 

 initiative of the investigator; he invents 

 hypotheses and he invents tests for them. 

 It is to the intellectual chai'acter of these 

 inventions that your attention is invited. 



The mental process by which hypotheses 

 are suggested is obscure. Ordinarily thej^ 

 flash into consciousness without premoni- 

 tion, and it would be easy to ascribe them 

 to a mysterious intuition or creative fac- 

 ulty ; but this would contravene one of the 

 broadest generalizations of modern psy- 

 chology. Just as in the domain of matter 

 nothing is created from nothing, just as in 

 the domain of life there is no spontaneous 

 generation, so in the domain of mind there 

 are no ideas which do not owe their exist- 

 ence to antecedent ideas which stand in the 

 relation of parent to child. It is only be- 

 cause our mental processes are largely con- 

 ducted outside the field of consciousness 

 that the lineage of ideas is diflicult to trace. 



To explain the origin of hypotheses I 

 have a hypothesis to present, — not, indeed, 

 as original, for it has been at least tacitly 

 assumed by various writers on scientific 

 method, but rather as worthy of more gen- 

 eral attention and recognition. It is that 

 hypotheses are always suggested through 

 analogj\ The unexplained phenomenon on 

 which the student fixes his attention re- 

 sembles in some of its features another 

 phenomenon of which the explanation is 

 know^n. Analogic reasoning suggests that 

 the desired explanation is similar in char- 



