DECLINE OF DETERMINISM EDDINGTON 149 



circles we must abolish purely retrospective characteristics — those 

 which are never found as existing but always as having existed. If 

 they do not manifest themselves until the moment that they cease 

 to exist, they can never be used for prediction except by those who 

 prophesy after the event. 



Chemical constitution is not a retrospective character though it is 

 often inferred retrospectively. The fact that silver nitrate can be 

 bought and sold shows that there is a property of hcing silver nitrate 

 as well as of having been silver nitrate. Apart from special methods 

 of determining the constitution or properties of a substance without 

 destroying it, there is one general method widely apjjlicable. We 

 divide the specimen into two parts, analyze one part (destroying it 

 if necessary) and show that its constitution has been X; then it is 

 usually a fair inference that the constitution of the other part is X. 

 It is sometimes argued that in this way a character inferable retro- 

 spectively must always be also inferable contemporaneously; if that 

 were true it would remove all danger of using retrospective infer- 

 ence to invent fictitious characters as causes of the events observed. 

 Actually the danger arises just at the point where the method of 

 sampling breaks down, viz, when we are concerned with character- 

 istics supposed to distinguish one individual atom from another atom 

 of the same substance; for the individual atom can not be divided 

 into two samples, one to analyze and one to preserve. Let us take 

 an example. 



It is known that potassium consists of two kinds of atoms, one 

 kind being radioactive and the other inert. Let us call the two kinds 

 Ka and K^. If we observe that a particular atom bursts in the 

 radioactive manner we shall infer that it was a K atom. Can we 

 say that the explosion was predetermined by the fact that it was a 

 Ka and not a K^ atom? On the information stated there is no 

 justification at all ; Ka is merely an antedated label which we attach 

 to the atom when we see that it has burst. We can always do that 

 however undetermined the event may be which occasions the label. 

 Actually, however, there is more information which shows that the 

 burst is not undetermined. Potassium is found to consist of two 

 isotopes of atomic weights 39 and 41; and it is believed that 41 is 

 the radioactive kind, 39 being inert. It is possible to separate the 

 two isotopes and to pick out atoms known to be K^x- Thus K^^ is 

 a contemporaneous character and can legitimately predetermine the 

 subsequent radioactive outburst; it replaces K^ which was a 

 retrospective character. 



So much for the fact of outburst; now consider the time of out- 

 burst. Nothing is known as to the time when a particular Ki^ atom 

 will burst except that it will probably be within the next thousand 



