THEORIES OF SCIENTIFIC CONCEPTS 163 



Poincare's position. If we have two bodies A and B, we 

 may formulate their relations in a law. But these relations 

 will be extremely complicated, hence we are led to suppose 

 two ideal bodies A' and B' whose ideal relations will be 

 described by the principles of geometry. 'The relation 

 between A and B was a rough law, and was broken up; 

 we now have two laws which express the relations of A 

 and A', of B and B\ and a principle which expresses that 

 of A' and B'" l But the principle is absolutely true in the 

 intensional sense, because it has been made so by definition, 

 i.e., we define A' and B' in such a way as to make the prin- 

 ciple necessarily true. But the principle cannot be true in 

 the empirical or extensional sense, for it is not applicable 

 to anything empirical. Hence we describe it as con- 

 venient. 



The net result of Poincare's view is somewhat as follows: 

 There are to be found in science symbols of a peculiar sort 

 called "conventions." These are not merely classificatory 

 in Pearson's sense, nor are they simply abstractions in 

 Hobson's sense. On the contrary, they are derived by elab- 

 orately creative processes which free them almost completely 

 from the data. Thus, in contrast with strict positivism, 

 the symbols have their meanings determined primarily by 

 the knowing operations and only incidentally by the data. 

 The creative operations, however, are performed with a 

 definite end in view; the symbols thus produced must be 

 convenient for science as a whole, i.e., they must be such 

 as to permit valid predictions. Hence the symbols — which 

 are at this stage essentially "empty" of meaning — are oper- 

 ationally defined in such a way as to make the predictions 

 possible. The creative knowing processes are precisely these 

 operational definitions. The symbols, once defined in this 

 way, cease to be true or false in the empirical sense, and 

 become only convenient. " The scientific fact [the convention] 

 is only the crude fact [the data] translated into a convenient 

 language ." 2 



1 Ibid., p. 336. 2 Ibid., p. 330. 



