ix EXPERIENTIAL RECONSTRUCTION 157 



at first sight we seem in sense-experience to be in direct 

 contact with outer realities, and if it were so, we had here 

 so firm a basis of knowledge that the only question to be 

 discussed would be the method of building upon it. The 

 Berkeleyan criticism soon showed that the matter was not 

 so simple, but even so it left experience standing as so 

 much fact, though fact of an c internal ' psychological kind, 

 and the problem of knowledge was to understand how 

 thought so organised experience as to discover general 

 truth. The Kantian and post-Kantian analysis showed, 

 however, that c immediate ' experience could not be taken 

 as a simple datum. From the outset we assert, and though 

 in sensation or in any form of immediate consciousness 

 that which we assert may, in a certain sense, 1 be taken as 

 fact, this sense is only reached by an effort of abstraction. 

 Analysis showed that the c pure experience ' which seemed 

 a prime starting-point was an abstraction from which the 

 elements of reference that piece its parts together are 

 omitted. The true starting-point of knowledge is the 

 assertion which assigns an object a place in a permanent 

 order, whereby it enters into relation with other objects. 

 Yet and this is the paradox of knowledge this order 

 is itself built up by slow degrees and is certainly not an 

 object of thought until experience is far advanced. The 

 solution of the paradox is that the cognitive life of mind 

 is from the first a correlating activity which connects the 

 successive phases and weaves them into a plastic order 

 to which every new experience is referred. It is true that 

 the precise nature of the reference is determined by rela- 

 tions which are contained in the objects of experience when 

 experience is taken as a whole, but (a) since experience 

 comes in fragments, spread over time, to take it as a whole 

 is only possible for a mind which can correlate distinct 

 data, and (b) certain methods of correlation, viz. those 

 which involve generalisation are never c given,' but in- 

 volve assertions going beyond anything that can be 

 given. 



Thus the unit of knowledge is an assertion involving 

 the object in relations, and the c pure ' experience which 

 1 See below, and footnote on following page. 



