66 WHAT IS SCIENCE? 



suppose we will return now to the instance of density 

 that when we re-determined the density and found it 

 changed, we could not detect any change in any other 

 property of the substance and that we could not find 

 a substance which, resembling this substance in all other 

 respects, had the density found in the first experiment, 

 could we then maintain the invariability of the associa- 

 tion ? Well, it would doubtless be very awkward and 

 men of science would get into a fever of excitement, but 

 they could maintain their law. For the supposition that 

 nothing had changed between the two experiments is 

 impossible to realize ; the mere fact that a previous 

 experiment had been made and that the second experi- 

 ment had been made after the first is sufficient to make 

 some change between the two. Of course our usual 

 conception of a substance excludes the idea that such 

 changes a mere repetition of a measurement or the mere 

 lapse of time could change its properties and make it a 

 new substance ; we should have to alter our conception 

 of a substance. But that conception has been already 

 altered so greatly since it was taken over from common 

 sense that there would be no impossibility and no insuper- 

 able inconsistency in maintaining that, since w r e made 

 the first experiment, the substance on which we made it 

 had vanished from our ken and been replaced by some 

 other substance, which might naturally enough have a 

 different density. 



Indeed we should have to maintain something of the 

 kind, for, whatever we might do, the fact would remain 

 that we have observed two densities which cannot be those 

 of the same substance and cannot be asserted by the same 

 law. Either we must include the two observations 

 under different laws, or we must leave one (or both) 

 of them outside laws altogether. We adopt this last 

 alternative if we regard the first measurement simply 

 as a mistake ; a mistake is something that is excluded 



