TESTIMONY AND AUTHOEITY. 75 and his opportunities for ascertaining the fact. The more varied the powers, experience and knowledge brought to bear on any question the more likely is the judgment to be well-founded. Any assertion concurred in by persons of different training, habits, and point of view is likely to be accordant with a wider aggregate body of knowledge and experience than if made by one of them alone ; and is accordingly so much the more reliable than if it had no other support. It is essential, however, in this case that the variety should be in knowledge or experience relevant to the assertion. This is an important part of the independence of the wit- nesses postulated in the mathematical treatment of testi- mony, and is perhaps more difficult of attainment than is a sufficient conflict of interest to eliminate bias and insin- cerity. Unfavourable cases are those in which it is not possible to obtain the requisite variety. An example may be easily imagined. Suppose we had the evidence of a number of men as to the colour of some object ; and after- wards found out that they all worked in some industry productive of colour-blindness. The effect on their testi- mony would be much the same as though they had agreed to deceive, and we should naturally refuse to rely on their assertions even though we saw no appearance of collusion. The remaining condition of trustworthiness whose relation to corroboration we have yet to mention is that of accuracy of memory and expression, and this does not seem to be touched by concurrence of testimony. On this, however, as well as on another of the conditions, light may be thrown by ' internal evidence ' which has been adverted to as constituting another form of corroboration. Rigid and thoroughgoing consistency in all parts cogency of reasoning and absence of confusion raise a presumption of accuracy and care. Even an occasional lapse as regards these characters is of no great consequence as invalidating the remainder ; for wherever consistency and cogency do appear they are assuredly not the result of chance. In the same way also that inaccuracy may be disclosed by inconsistencies of statement, a presumption of imperfect observation may be raised by a record that fails to distinguish exactly between results due to observation and those due to- inference. Where these are not carefully discriminated in description and discussion there can be no certainty that they have not been confused in fact ; and accordingly the work in which they appear cannot be regarded as fulfilling the condition that has reference to observation. The third and last form of corroboration is that gained by
