DESCRIPTIVE TECHNIQUES 107 



by retrospective inference he could never identify objects 

 without destroying them, and they would reduce to mere 

 bundles of potentialities. The very fact that the observer 

 speaks of these operations as changes indicates that he is 

 able to become independently aware of the objects prior to 

 the changes. Practically, one avoids difficulties of this kind 

 by employing as materials upon which operations are per- 

 formed not all of the objects of a given kind, but merely 

 samples from them. In destroying the samples, therefore, 

 he still has left objects of the kind which have been destroyed, 

 and which can be characterized in terms of their capacities 

 to behave in this particular way when subjected to this 

 particular kind of operation. 



The manipulation of objects in this way constitutes what 

 is ordinarily called experimentation. But a rather sharp 

 distinction may be made between experiments for discovery 

 and experiments for verification. The type which is here 

 being considered belongs in the former category, but the 

 more usual sort belongs in the latter. Experimentation for 

 discovery belongs among the descriptive techniques, and is 

 analogous to the more or less random toying, pinching, 

 turning, and tasting activities which the child uses to explore 

 a new object. The operations are performed without specific 

 direction or control, and without any well-defined idea as 

 to what they may reveal. They are thus activities in which 

 one discovers something which he had not previously known 

 about the object. Verificatory experiments, on the other 

 hand, as will be seen in Chapter XI, although they involve 

 discovery, are definitely planned in advance, and involve a 

 more or less precise prediction that a certain outcome will 

 ensue. The latter presuppose that one has already learned 

 enough about nature to anticipate certain other of its fea- 

 tures ; the former presuppose nothing but the general capacity 

 of nature to reveal itself under different guises. 



There is some doubt as to whether changes in the location 

 of an object should be included among the controls of the 

 end-object or among the controls of the physical medium. 



