XXIII.] THE USE OF HYPOTHESIS. 511 



laws of nature, it is self- condemned. We cannot even 

 apply deductive reasoning to a self-contradictory notion ; 

 and being opposed to the most general and certain laws 

 known to us, the primary laws of thought, it thereby con- 

 spicuously fails to agree with facts. Since nature, again, 

 is never self-contradictory, we cannot at the same time 

 accept two theories which lead to contradictory results. 

 If the one agrees with nature, the other cannot. Hence if 

 there be a law which we believe with high probability to 

 be verified by observation, we must not fiamean hypothesis 

 in conflict with it, otherwise the hypothesis will necessarily 

 be in disagreement with observation. Since no law or 

 hypothesis is proved, indeed, with absolute certainty, there 

 is always a chance, however slight, that the new hypo- 

 thesis may displace the old one ; but the greater the pro- 

 bability which we assign to that old hypothesis, the greater 

 must be the evidence required in favour of the new and 

 conflicting one. 



I assert, then, that there is but one test of a good 

 hypothesis, namely, its conformity vnth observed facts ; but 

 this condition may be said to involve three constituent 

 conditions, nearly equivalent to those suggested by Hobbes 

 and Boyle, namely : — 



(i) That it allow of the application of deductive reason- 

 ing and the inference of consequences capable of com- 

 parison with the results of observation. 



(2) Tiiat it do not conflict with any laws of nature, or 

 of mind, which we hold to be true. 



(3) That the consequences inferred do agree with facts 

 of observation. 



Possibility of Deductive Reasoning. 



As the truth of an hypothesis is to be proved by its 

 conibrmity with fact, the first condition is that we be able 

 to apply methods of deductive reasoning, and learn what 

 should happen according to such an hypothesis. Even 

 if we could imagine an object aiiting according to laws 

 hitlierto wholly unixiiown it M'ould be useless to do so, 

 liecause we could never decide whether it existed or not. 

 We can only infer what would liappen under supposed 

 conditions by applying the knowledge of naiure we possess 



