494 THE ROYAL SOCIETY OF CANADA 



records or narratives, like the individual quantities in a mathematical 

 table, being the product of correct processes, are prima facie correct, 

 and ought to be accepted by any user (reader or subsequent historical 

 investigator) without further test, if such test be impossible or unreas- 

 onably inconvenient. This acceptance is for the same reason as in the 

 case of the table : to do otherwise would defeat the very purpose of the 

 record or narrative, and this is not necessary because such untested 

 use will reproduce in the user's data and own investigations and narra- 

 tive an average of essential correctness corresponding approximately 

 to the average of essential correctness in the records or narratives 

 so used. 



All that could ever be proved against such records, narratives 

 and tables, and against the results of such users, would be a proportion 

 of incidental error not sufificient to lower their average of correctness 

 beneath the required average of the respective branches of science in- 

 volved. Under the application of correct processes, the average of es- 

 sential correctness in history, though not equal to the mathematical, 

 exceeds that of medical practice and of the courts, because physicians 

 and courts are obliged to make a decision even if the available data 

 and evidence are unsatisfactory, but the historian, when in doubt, can 

 always preserve silence. Under the prevailing method, however, 

 history is the great exception among all the sciences. It has no funda- 

 mental correct process or processes. The nearest approach to one is 

 its test for certainty by agreement of two or more independent sources 

 in confirmation of any specific point. According to Seignobos, how- 

 ever, who is one of the two more prominent recent writers on the 

 prevailing method, this confirmation by two or more records is most 

 frequently lacking save in recent history; and even where the con- 

 firmation is available the original independence of the confirming 

 records cannot be conclusively shown except in modern periods.^ 

 Bernheim, the other writer, believes that this estimate of the number 

 of confirmed points is too small, but one need not depend on the 

 opinion of either: it is clear in itself what a blanketing and disruption 

 of history would follow if all singly attested periods and points were 

 stricken from its reckoning. The most important single features in 

 any historical method are its test for trustworthiness in records and 

 its treatment of conflicting statements in trustworthy records where 

 the circumstances of the discrepancy are unknown. In the prevailing 

 method the treatment of such discrepancies is to attempt a harmoniza- 



^ Langlois & Seignobos, Introduction aux Etudes Historiques, pp. 168, 174. 

 Seignobos wrote the part of the book containing the passages cited. He is professor 

 of contemporary history at the Sorbonne (University of Paris). Bernheim is professor 

 of history at Greifswald in the Prussian province of Pomerania. 



