PREFACE TO THE FIKST EDITION. 



accurate observation, for successful experiment, and for 

 the sure detection of the quantitative laws of nature. 

 As it is impossible to comprehend aright the value of 

 quantitative laws without constantly bearing in inind the 

 degree of quantitative approximation to the truth probably 

 attained, I have devoted a special chapter to the Theory 

 of Approximation, and however imperfectly I may have 

 treated this subject, I must look upon it as a very essential 

 part of a work on Scientific Method. 



It then remains to illustrate the sound use of hypo- 

 thesis, to distinguish between the portions of knowledge 

 which we owe to empirical observation, to accidental dis- 

 covery, or to scientific prediction. Interesting questions 

 arise concerning the accordance of quantitative theories 

 and experiments, and I point out how the successive 

 verification of an hypothesis by distinct methods of ex- 

 periment yields conclusions approximating to but never 

 attaining certainty. Additional illustrations of the general 

 procedure of inductive investigations are given in a 

 chapter on the Character of the Experimentalist, in which 

 I endeavour to show, moreover, that the inverse use of 

 deduction was really the logical method of such great 

 masters of experimental inquiry as Newton, Huyghens, 

 and Faraday. 



In treating Generalisation and Analogy, I consider the 

 precautions requisite in inferring from one case to another, 

 or from one part of the universe to another part ; the 

 validity of all such inferences resting ultimately upon 

 the inverse method of probabilities. The treatment of 

 Exceptional Phenomena appeared to afford an interesting 

 subject for a further chapter illustrating the various modes 

 in which an outstanding fact may eventually be explained. 

 The formal part of the book closes with the subject of 

 Classification, which is, however, very inadequately treated. 

 I have, in fact, almost restricted myself to showing that 

 all classification is fundamentally carried out upon the 



