THE DISCOVERY OF LA\Y> 63 



false, for some swans in Australia are black. Instances 

 of the second kind are plentiful in actual science. When 

 a chemist makes a new compound, he often determines 

 its melting-point or density ; as a result of a single 

 measurement he will often be prepared to assert that 

 its melting-point or density is higher (say) than that of 

 water, and nobody will dream of doubting that the asso- 

 ciation he asserts is invariable or that subsequent measure- 

 ments will lead to the same result. 



These examples seem to prove that a large number of 

 favourable instances, even if without exceptions, is neither 

 sufficient nor necessary to establish a law. But at the 

 same time they suggest what is the additional element re- 

 quired. We have omitted to take into consideration other 



closely similar to those that are under discussion. 

 The chemist is certain that, in measuring the melting- 

 point of a new compound, he is establishing an invariable 

 relation, because from the examination of a great number 

 of other compounds he has found that the density is an 

 invariable property. On the other hand seventeenth- 

 century naturalists ought to have regarded with suspicion 

 a law that all swans are white (and probably they did 



illy so regard it) because the examination of other 

 animals would have shown them that colour is by no 

 means an invariable property, but is liable to vary very 



ly even among closely related species. In putting 



the matter as we did, the full evidence was not disclosed. 



The evidence for the invariable density of a new compound 



ingle measurement of it, but the general law 



all densities are invariable properties. This law is 



>lished by the observation, not of a single instance 

 or of one or two, but of a very large number of instai 

 in none of which tin,- n 1 it ion has been found to fail. 

 o for the assertion of the law of the density 

 of the n ance is really of exactly the same nature 



as that for t an to-morrow. 



