100 INTRODUCTION TO PHILOSOPHY OF SCIENCE 



magnetic and gravitational phenomena are present in as- 

 certainable amounts and kinds. Then a normal observa- 

 tional situation would be one which is psychologically, 

 biologically, and physically normal. 



The possibility of determining laws of operators is then 

 based upon a practical assumption: in normal observation 

 the transforming effect of the operators may be neglected. This 

 is equivalent to saying that in this type of situation the 

 operators are considered as non-transforming. On this as- 

 sumption one is able to become "directly' aware of end- 

 objects, pointer readings, and sense-data. He may therefore 

 compare end-objects with pointer readings and by this 

 means determine the laws of P-operators; he may compare 

 pointer readings with sense-data and thus determine the 

 laws of B-operators; he may compare sense-data with " in- 

 terpretations'* and, as a result, formulate the laws of 

 M-operators. He then sees that his assumption with refer- 

 ence to the non-transforming power of operators represents 

 a first approximation, and he corrects it by means of the 

 laws already established. This then becomes a second 

 approximation, and he corrects the laws of the individual 

 operators accordingly. By this means he reaches a progres- 

 sively more accurate and reliable knowledge of the laws of 

 nature, and feels himself justified in using them as principles 

 of inference. 



The second feature of inference which must be called to 

 attention has to do with its relation to symbols. Inference 

 functions in observation in affording a technique on the 

 basis of which from a knowledge of an object in a certain 

 context one may determine its character in another. How- 

 ever, although the appearance of which one is directly 

 aware is the basis for an inference to the character of the 

 object in a different context, the appearance should not 

 properly be called the symbol of the inferred object. The 

 greatest confusion is introduced into Eddington's position 

 by this terminology. For example, he fails to make a distinc- 

 tion between pointer readings — which are clearly events in 



