QUANTITATIVE ESTIMATES OF SENSORY EVENTS 319 



relations, including those involving absolute length, are the same. But 

 the two relation structures are not identical ; they are separate instances of 

 the same relations. If we regard three balls arranged as above as a 

 phenomenal unit — a ' thing ' — the two relation structures are not the same 

 thing but two different things, each with its own identity, even if there is 

 no possible test by which we can distinguish one from the other, as there 

 will not be if all the relations in the two structures are identical. In 

 describing phenomena as relation structures we do not mean a con- 

 glomeration of abstract relations, but a grouping of actual instances of 

 relations. Some modern philosophers are apt to forget this and regard 

 the universe as built up of abstract mathematical material. But we cannot 

 build reality with abstractions : there must be ' things ' to relate before 

 we can have an example of any relation. It is true that these ' things ' are 

 themselves resolvable into relation structures involving relations between 

 more elementary ' things,' and so on indefinitely, but we never reach a 

 stage at which we find relations without ' things ' to relate. This is not 

 because methods of physical analysis do not go far enough. If our present 

 ultimate things — electrons, positrons, etc. — were split into a million other 

 things, and each of these into a million others, we should still be in the 

 same position. Any actual instance of a relation must have things in it to 

 be related. All the relations in the relation structures constituting two 

 billiard balls may be identical, but the relation structures are not identical ; 

 the ' things ' in them are different : it requires twice as many of them to 

 make two balls as to make one. Therefore when we say that two objects 

 have the same relation structure we do not imply identity. This is clear 

 enough when we are considering two or more objects like billiard balls 

 whose non-identity can be proved by their simultaneous existence in 

 different places. It is not so obvious in the case of the relation structures 

 of which the same object is composed at different times. Here the 

 elementary ' things ' which enter into the various relations in the structure 

 are the same things on all occasions, or at any rate we have no way of 

 knowing that they are not. But we have no way of knowing that they 

 are. Our idea of a permanent object is based on the inter-consistency of 

 an enormous number of acts of perception involving different aspects of 

 the object and different relations to other objects. In any single act of 

 observation only a small part of this material is presented to us, and its 

 perception is a self-contained experience. We perceive a phenomenal 

 relation structure, that is all. On a subsequent occasion we also perceive 

 a phenomenal relation structure. How do we correlate these experiences ? 

 Neither of them in itself contains anything to indicate that it is a repetition 

 of the ' same ' experience. As far as immediate perception is concerned 

 we must regard the relation structures perceived on separate occasions as 

 separate structures, and we can only correlate them by means of cross 

 relations between these structures. The correlation of successive appear- 

 ances of the same object therefore depends on the same principle as the 

 correlation of the appearances of two or more objects simultaneously 

 observed. In the latter case the relation structures are separated in the 

 dimension of space and in the former in the dimension of time, but for 

 the comparison of perceived relations this difference in the dimension of 

 separation is of no importance. Perceptions separated in time are com- 

 parable in consequence of memory which reproduces the relation structure 

 of a past observation for comparison with the relation structure of a present 

 one. The smaller the gap in time between the observations, the more 

 accurately is the memorised structure reproduced. As a matter of fact 

 the comparison of the relation structures of simultaneously existing objects 



