216 ANNUAL REPORT SMITHSONIAN INSTITUTION, 19 3 8 



unobservable by any means, existing or imaginable, would be logi- 

 cally unobservable. Anything which would be observable by the 

 existing means if we were also omnipotent, but which actually is 

 unobservable because we cannot make full use of the means of 

 observation which exist, would be practically unobservable. 



We can now proceed a stage further. We have just seen that, for 

 our purpose, the humanly and the physically unobservable become 

 one class because we cannot at present distinguish between them. 

 Let us look, then, at our latest classification, to make sure that the 

 three types of unobservability we have now reached are immediately 

 distinguishable. It is fairly evident, I think, that they are, if we grant 

 the initial supposition that we have discovered all existing means of 

 observation. For brevity, I will call that the assumption of omnis- 

 cience, and you will understand that by that word I do not mean 

 knowledge of everything that exists or that can be observed, but 

 complete knowledge of the existence and properties of every means 

 by which observation is possible. For example, omniscience implies 

 complete knowledge of all the properties of light, but not necessarily 

 of all objects which are visible. Now, clearly, this assumption of 

 omniscience is open to challenge, and it is, therefore, necessary to 

 see how our classification looks if it is removed. Can we then still 

 recognize the three classes as distinct? 



There is no difficulty, I think, with the logically unobservable. 

 Since this class consists of things which are not even imaginably 

 observable, it makes no difference how much we know of possible 

 observability. There can be no possible means of observation that 

 is not imaginable. There may be unimagined means, of course. We 

 may make discoveries that take us by surprise, but those discoveries 

 must have been imaginable, or we could not have apprehended them 

 at all. Whatever we know of observability, then, does not affect our 

 potentialities of observation, so that the logically unobservable is a 

 definitely distinguishable class, independently of our assumption of 

 omniscience. 



The case is different, however, with the practically and the physi- 

 cally unobservable, for these classes cannot be distinguished if we do 

 not regard ourselves as omniscient. We said that the far side of the 

 moon was practically unobservable, but if we are not omniscient, 

 how do we know that when we have overcome what seem to be the 

 present difficulties; when we have made vessels with adequate air and 

 food supply, and vehicles that we can drive accurately and swiftly 

 enough to take us unharmed to the moon during a week-end — how 

 do we know that nature will not then face us with some unexpected 

 difficulty like that which she kept in store for our efforts to determine 

 our motion through the ether? If she does, and persists in doing the 

 same kind of thing, we shall have to call the far side of the moon 



