THE DISCOVERY OF LAWS 05 



instance must destroy the law. For an. association which 

 has failed once, even if it has not failed a million times, 

 is not strictly invariable. True ; but what exactly is 

 the association we are asserting ? We are asserting that 

 a certain density is invariably associated with a certain 

 substance. If we find a new density we cannot maintain 

 the invariable association if we attribute it to the same 

 substance as that to which the old density was attributed. 

 But why should we not attribute it to a new substance ? 

 If we try the experiment over again and find that we do 

 not get the same result as before, what is to prevent us 

 avoiding any discrepancy between the two experiments 

 by simply saying that they are not made on the same 

 substance ? 



Indeed this way out of the difficulty has been adopted 

 implicitly in the case of the black swan. Since we have 

 known of black swans, we do not say that there are not 

 white swans ; we recognize two kinds of swans, one of 

 which is black and the other white. Nor do we recognize 

 any error in the assertion, by those who did not know of 

 black swans, that all swans are white. All the swans 

 that they knew anything about were white and have 

 always remained white. The apparent difficulty arose 

 only because the new birds were called swans. If we 

 confine that term to the birds which were originally called 

 swans, any law about swans is quite unaffected by the 

 discovery of birds which resemble swans in some respects, 

 but which, since they are not wholly the same, should 

 not be called swans. 



But, it may be urged, the case is not really parallel to 

 that which we must suppose if we want to face the 

 difficulty fairly. Black swans differ from white in other 

 things than their colour, so that there is a reason quite 

 apart from their unexpected colour for distinguishing 

 them from white swans. Again, even after the discovery 

 of black swans, white swans could still be found. But 



