538 THE ROYAL SOCIETY OF CANADA 



also experimentally the principle that in history, as in other sciences, 

 probable conclusions even of the superior measure are without signifi- 

 cance except where they represent averaged results; and historical 

 probable conclusions, in so far as that term implies the mere balancing 

 of reasons pro and con, or inferior form of probability, are condemned 

 without exception as without significance and unscientific because 

 this form of probability cannot be averaged and has even less signifi- 

 cance than the superior in producing individually correct results. 



b. General features of the erratic nature of probability. The erratic 

 nature of probability even in its superior form is shown not only by 

 the continual displacement and final elimination of the 41 conclusions 

 as a result of additional evidence, but also by the random manner in 

 which the degrees of probability are found distributed in the successive 

 conclusions of each case. There is no uniform increase in the prob- 

 ability of the conclusions as the amount of evidence increases, but the 

 degree of probability in the successive conclusions rises in some in- 

 stances and falls in others, and sometimes remains the same. Another 

 general instance of the erratic nature of probability is afforded by 

 probable harmonizations of conflicting statements. Where trust- 

 worthy records conflict with one another and the circumstances of 

 the discrepancy are unknown, it is a common practice to attempt 

 a probable harmonization. In order to test the value of this practice 

 experimentally, the author in 1909-1911 gathered from actual inter- 

 course 26 cases in which statements that were conflicting were yet 

 true. On the assumption that the conflicting statements in the 26 

 cases were recorded without mention of the attendant circumstances, 

 it was found in 14 of the cases that a probable harmonization of the 

 statements seemed easy but in the others impossible. There was 

 nothing in the attendant circumstances of the discrepancies to account 

 for this erratic division into two groups under an identical test. In 

 the 14 cases, moreover, where a probable harmonization seemed easy, 

 the harmonization proved in every instance to be contrary to the 

 facts. The details of this test in the 26 cases are given in a paper on 

 the Origin and Treatment of Discrepancy in Trustworthy Records 

 in the Transactions of the Society, 1911, vol. v, 127-178. Upwards 

 of 150 further cases subsequently located by the author in actual 

 intercourse confirm the results shown by the original 26 under similar 

 tests. 



c. Special features of the erratic nature of probability. Two special 

 features of the erratic nature of probability are illustrated by Cases 

 8 and 12 in the present test, as follows: — 



1. In Case 8, if it be placed in evidence that there are two daugh- 

 ters with difi^erent surnames and nothing further is specified except 



