THE USES OF EXPERIMENT 43 



by which we test the truth of our inductions. Obviously, we 

 cannot test the truth of either of the foregoing inductions, nor 

 decide between them, unless we appeal deductively back again to 

 the facts of the world around us. Experiment is one of the ways 

 in which we may test an inference. After we have considered a 

 collection of facts and drawn an induction, we say to ourselves, 

 " If our hypothesis is true, if we have thought correctly, such and 

 such consequences must follow " ; and, if possible, we experiment to 

 ascertain if they do follow. If they do actually follow, then there 

 is a high probability that our hypothesis is correct, and that 

 probability is raised to a higher degree by every new and success- 

 ful appeal to reality. On the other hand, a valid appeal, experi- 

 mental or other, if unfavourable, immediately and conclusively 

 demonstrates the falsity of the hypothesis, which therefore must be 

 abandoned or emended. " Thus Kepler records that he advanced 

 nineteen hypotheses which he afterwards disproved, before he 

 arrived at the true statement of the laws of planetary motion." l 

 " A similar spirit was shown by Newton in respect to his 

 hypothesis that the moon is retained in her orbit by the force of 

 gravity. From this hypothesis he calculated that the moon ought 

 to be deflected from the tangent of its orbit something more than 

 fifteen feet every minute. But the apparent deflection was only 

 thirteen feet. This discrepancy, comparatively small though it was, 

 Newton accepted as a disproof of his hypothesis, and ' laid aside at 

 that time any further thoughts of this matter.' But some fifteen 

 years later the distance of the moon from the earth had been more 

 exactly ascertained, and Newton repeated his calculations, working 

 with these new values. The agreement between the calculated 

 and the normal deflection was then seen to be remarkably precise, 

 and the hypothesis became an established theory." 2 



73. Unless an experiment is entirely aimless, entirely a product 

 of unscientific thought, it is always an outcome of deduction. 

 Even when it is used, not to test thinking, but merely to elicit 

 fresh information, it is still an outcome of deduction ; for it is de- 

 duction which leads us to suppose that our experiment will have 

 useful results. It follows that there are two ways in which we 

 may use experiment (or any other laboratory method). First we 

 may use it as a deductive test of previous thinking, in which it 

 tends to secure great accuracy in thinking. Second, we may use 

 it as a means to reveal an obscured fact, when, of course, it tends 

 to secure accuracy in observing. But, in the latter case, if we base 

 1 Welton, Manual of Logic, vol. ii. p. 66. 2 Op. cit., vol. ii. p. 88. 



