ON THE STATISTICAL VIEW OF NATURE. 587 



these generalisations can be tested, elevated to the rank 

 of leading canons of thought and research, and in rare 

 cases to that of the expression of a law of nature. So 

 far, therefore, as the complicated phenomena presented 

 in meteorology, agriculture, and economics are concerned, 

 . the suggestions leading to so-called laws have in every 

 case been got elsewhere — from astronomy, chemistry, 

 psychology, history, &c. ; and the work of science has 

 subsequently consisted largely in gathering the necessary 

 statistical materials by which to prove, amplify, curtail, 

 or refute them. In many cases it has been found that 

 even elaborate series of observations had not been per- 

 formed in such a manner ^ as would permit of the 

 necessary inferences being drawn from them. Similarly 

 biologists after Darwin have had to rearrange the collec- 

 tions made by those who came before the epoch marked 

 by that great name. 



^ This refers as much to statist- 

 ical figures as to the knowledge 

 accumulated in many of the natural 

 sciences. Especially it refers to the 

 statistical material upon which 

 Quetelet based his st;irtling and 

 epix-h-making assertions : theearlier 

 critics had, as V. John observes 

 (' Geschichte der St;itistik,' p. 364), 

 dealt with the deductions whicii 

 (Quetelet had drawn, without deal- 

 ing witii the empirical material 

 itself. It was therefore of great 

 importance that Prof. Kchnisch of 

 Gottingen for the first time sub- 

 mitted the figui'es themselves to a 

 searching analysis. He did this 

 in the years 1875-76, in his articles 

 in the ' Zeitschrift fiir I'hilosophie 

 und Philosophische Kritik,' through 

 which it became evident that the 

 inferences were, as Lotze had 

 already suggested, to sny the least. 



premature. " In the memoir ' Sur 

 le Penchant au Crime' (1831), only 

 four years, and in the work ' Sur 

 THomme,' only six years (1826-31) 

 of the 'compte general,' furnished 

 the data upon which the astound- 

 ing regularity with which crime 

 repeats itself was maintained " (V. 

 John, p. 365). Rehnisch adds many 

 other examples of the extreme in- 

 completeness of the recoids upon 

 which the theory of (Quetelet is 

 built up. More recent labours 

 have therefore been to a large 

 extent directed towards gathering 

 inoie complete statistical data, as 

 well as towards improving the 

 mathematical methods themselves 

 to which not only these but also 

 the population and mortality 

 statistics have been submitted, for 

 the purpose of arriving at average 

 figures. 



