140 GENERAL VIEW AND BASIS OF SCIENTIFIC RESEARCH. 



nitude whilst varying the magnitude of the true pheno- 

 menon, or make it vary at a different rate ; and in cases 

 where it is small and cannot be rendered constant in 

 amount, we ascertain its mean value, and make an allow- 

 ance for it. We must not, however, ' cook the accounts,' 

 nor must we include single large errors when taking the 

 mean, because they may be due to some unperceived and 

 special circumstance. We must also not exclude divergent 

 results, because a divergent result may really be an im- 

 portant one by being an instance of a new class of facts, 

 or of a new cause or law, and require a new research. 

 Errors of excess of effect are as often likely as those of 

 deficiency. 



Even after the most perfect investigation, a number of 

 errors must remain, because we perceive only a minute 

 fraction of the phenomena which are involved in and 

 related to the substance or action which we have ex- 

 amined. In some subjects also, for instance, the complex 

 and concrete sciences, many errors must exist, because 

 they are vague ; many also must exist, because they can- 

 not be disproved. Such errors are best allowed to perish 

 by neglect. The most impregnable and lasting errors, 

 and which do not perish by neglect (because they are con- 

 tinually revived by fresh generations of believers), are 

 those which flatter the weakness of mankind, which are 

 continually being impressed upon us, and which are rarely 

 contradicted, because they cannot be individually dis- 

 proved, as well as because of the personal risk of question- 

 ing the truthfulness of any favourite popular belief. Such 

 errors are known to be such, only by means of the funda- 

 mental logical principle, that if two statements contradict 

 each other, one must be erroneous, and both may be so : 

 the difficulty really consists in proving in which statement 

 the error lies. 



