Essays in Statistics. 191 



but only at averages. We do not find in his book anything like an 

 analytical research into the general formulas of heredity. His 

 method is statistical. And here the question arises, What is the 

 value of this method, applied to moral facts ? 



Statistics, according to the definition of its professors, is ' the 

 science of social facts expressed in numerical terms.' Its object 

 is to collect and group methodically all moral or social phenomena 

 which are susceptible of numerical valuation. Its method consists 

 in exposition and induction. The method of exposition, which is 

 the simple and the more certain, consists in the calculation of 

 averages, and is based on this undoubted truth, that ' in an inde- 

 finitely protracted series of events, the action of regular and con- 

 stant causes must in the long run outweigh that of irregular causes.' 

 (Laplace). The inductive method, which is less certain, consists 

 in obtaining numerical expressions for social facts, by means of 

 arithmetical or algebraic processes applied to a small number of 

 observations, and in admitting, on the ground of analogy or prob- 

 ability, results not directly established. Mr. Gallon employs both 

 methods, but chiefly the second. He feels, therefore, confident in 

 regard to his method. 



In spite of all the attacks and jokes levelled against it, I hold 

 that statistics is a genuine science, and that it is of high importance. 

 But its mistake, in my opinion, is to suppose that it furnishes a 

 quantitative determination. As we have seen, science has two 

 chief phases : the one where it takes its rise in becoming objec- 

 tive; the other where it attains its perfect form in becoming 

 quantitative. Statistics halts at the first, while thinking to reach 

 the second. 



To see that this is so, in spite of appearances, in spite of columns 

 of figures and the imposing array of calculations, we will take a 

 moral and social fact of high importance human liberty. An 

 attempt has been made to study it by means of statistical data. 

 Quelelet in his Physique Sociale, and after him Buckle in his His- 

 tory of Civilization, have used these with great ability. They have 

 shown that the amount of crime in general, and of each species of 

 crime in particular, varies much less than is supposed ; that in the 

 beginning of each year, supposing the circumstances to remain the 

 same, we might almost predict with certainty the numb3r of crimes 



