OF TESTIMONIES OR JUDGMENTS. 603 



But it is usual, in solving these problems, to regard such events as com- 

 pound, and to derive them from a hypothesis which presents as its scheme or 

 system of data, the probabilities of individual correctness of judgment in the 

 members of the jury ; the correctness of judgment in any such members being 

 regarded as a simple event. And this mode of procedure is a very natural and 

 very obvious one. For the degree of unanimity of a decision will so far depend 

 upon the correctness of judgment in the members, that, if we knew what the 

 probability of correctness in each member was, we could determine a priori the 

 probability of any proposed measure of agreement in the body. 



The only question which arises, indeed, is not concerning the necessity of the 

 postulate, but concerning the mode in which it may be lawfully applied. How 

 shall we lawfully construct the hypothesis by which the solution of a problem 

 shall be made to depend upon the consideration of simple events. In answering 

 this question, I will endeavour to show, 1st, upon what the construction of the 

 hypothesis does not depend ; 2dly, upon what it does depend. 



11. The legitimate construction of the hypothesis in question cannot depend 

 upon the accidents of language, or causes deeper than accident, which have led 

 us to express particular things or events by simple terms, thus regarding them 

 as simple events ; and other events by combinations of these simple terms, thus 

 presenting them as compound. The solution of a question in the theory of proba- 

 bilities must depend upon the information conveyed in the data, not upon the 

 peculiar elements and constructions of the language which is the vehicle of that 

 information. Languages differ widely in these respects. Objects and events 

 which in one language are expressed by simple terms, are in another expressed 

 by combinations of simple terms. It is affirmed that a perfectly general method 

 of solution must be independent of, and superior to, differences like these. 



I will endeavour to illustrate this principle by an example. Let the problem 

 to be resolved be the following. The probability of the concurrence of rain and 

 snow is p, of the concurrence of snow and wind q and of the concurrence of wind 

 and rain r ; required the probability of the concurrence of wind, rain, and snow. 



Now suppose that we had to interpret the problem into a language in which 

 there were no simple terms corresponding to the simple terms "wind," "rain," 

 " snow," but in which there were simple terms for the three first of the concur- 

 rences above described. 



We may, for simplicity, suppose that language to be a dialect of English, and 

 the concurrence of rain and snow to be represented in it by the term " sleet," 

 the concurrence of snow and wind by the term "drift," and the concurrence of 

 wind and rain by the term "storm." 



The event whose probability is sought, viz., the concurrence of rain, snow, and 

 wind, would, in such a language, be represented by the combination either of two 

 of the terms above defined (as of sleet with drift), or of all the terms together, 



