DISCOVERY BY MAKING CONVERSE EXPERIMENTS. 549 



thermo-electricity. In devising experiments of this kind, 

 the great mechanical law of equality of action and reaction, 

 and the principle of equivalency of forces, must be re- 

 membered ; also the fact, that whilst in a direct experi- 

 ment a cause may produce a conspicuous effect, in the 

 converse experiment the causes may be distributed, or so 

 operate as to produce only a feeble effect ; and the degree 

 of any particular effect cannot usually be determined 

 beforehand. The converse form of Faraday's great dis- 

 covery of the magnetic rotation of polarised light has not yet 

 been discovered, and is well worthy of being sought for. 



e. By subjecting different forces or a series of sub- 

 stances to similar new conditions. This method is 

 adapted to discover the degree of universality of a particular 

 property. For instance, by suspending a great variety of 

 substances free to move between the poles of a powerful 

 magnet, Faraday discovered the universality of magnetism ; 

 Coulomb had previously made a similar attempt. 



It is also eminently suitable for discovering entirely 

 new and unsuspected phenomena. By it we may solve 

 the singular and apparently insoluble problem, how to 

 search for something neiv, and be sure of finding it. 

 This method, like most others, depends for its success 

 upon the principle, that if we put matter or its forces 

 under new conditions, new effects take place. If we 

 further put a sufficient number of forces or substances 

 under the new conditions, some of the more conspicuous 

 instances of the new effect are very likely to be in- 

 cluded, and, being conspicuous, will be thrust upon our 

 notice. This plan usually affords only a very small pro- 

 portion (I have found about one or two per cent.) of new 

 results ; it is, however, a certain one, provided one or two 

 hundred substances are employed. But if only a small 

 number of substances are used it is very uncertain, because 



