Fundamental Principles of Scientific Inquiry. ?>11 



those obtained by conventional interpolation. The status 

 of simplicity in scientific method, and in particular its 

 relation to probability, therefore requires examination. 



We notice that where a simple law exists it is obtained 

 from the observations ; then it is found that the obser- 

 vations do not fit the law exactly. The discrepancies are 

 called errors, and usually ignored afterwards. This is 

 so far conventional, but nevertheless the convention is an 

 extremely important one. The use of the word " error " 

 has led to the idea that there is something wrong with 

 such observations, which is not the case. The observation 

 is always right, in consequence of the mere fact that it 

 is an observation. All errors of observation seem to 1 e 

 reducible to this conventional form : a physical law is 

 chosen to fit the observations as closely as possible, giving, 

 in other words, a first approximation ; then another law 

 is discovered to be relevant, and the two together give a 

 closer approximation, and so on, the error at each stage 

 being merely the unexplained balance. Thus an error is 

 just the difference between an observed quantity and its 

 value predicted by the combination of all the physical laws 

 so far known to be relevant. The "theory of errors" 

 appears to be an extraneous hypothesis, based on a parti- 

 cular plausible assumption about the nature of errors; 

 its utility is chiefly in giving a unique solution in particular 

 cases where the solutions that fit the facts almost equally well 

 differ considerably among themselves. This advantage is, 

 however, liable to be exaggerated *. 



It is also necessary to accept testimony ; some primitive 

 propositions may be involved in this, but it is certainly a 

 practice that requires further analysis. It is often stated't 

 to be based on certain resemblances of behaviour between 

 the subject and other people, the subject knowing what 

 he would mean if he made the statements he hears or reads. 

 This is probably largely true, but it is not the way in which 

 one actually arrives at the principle in the first place ; 

 a child obeys orders and otherwise shows by its behaviour 

 that it can understand speech long before it is itself able 

 to speak. The relation between other persons'' speech and 

 the events that it represents is therefore originally direct 

 and not by way of the subject's own speech. Tic could 

 accept testimony just as well it: he had been dumb all his life. 



* N. r. Campbell, 'Physics: The Elements,' 1920, p. 507. 



f Karl Pearson and others base the belief in the consciousness of 

 other persons on such resemblances; but testimony appears to be 

 -acceptable without the introduction of other persons* consciousness. 



