Everybody normally acts on the assumption that the value of human testimony is an extremely variable quantity. The rules by which we assess the value of testimony, in the ordinary affairs of life, are of that thoroughly habitual kind that hardly involve conscious processes; they repose on two judgments, which we are always making. Our belief in direct testimony to an event is conditioned by the nature of the event and by our estimate of the “personal equation” of the witness. These two factors are not quite independent; it is very seldom, for instance, that we attribute “general untrustworthiness” to anybody who is known to us. Our experience usually teaches us that there are certain classes of statements--e.g. his breaks at billiards, the number of miles his motor-car runs on a gallon of petrol--for which that particular witness’s credibility is at a minimum. For some other classes of statements we may have learnt to take his word without hesitation. What the mathematicians call the “credibility” of a witness is not, in the case of any witness of whom we have personal knowledge, a constant figure. It varies with the event, often in an extremely complicated way. When Brown is listening to Jones talking about the enormities of Smith the extremely delicate and rapid weighing of probabilities being performed by Brown beggars any mathematical description. When the witness is personally unknown to us the matter becomes simpler. Our conclusions, one way or another, will be held with less confidence, but they will be more simply arrived at. We may, on the evidence supplied by the testimony itself--the tone of the letter, the man’s manner in the dock--class the witness as a man of a certain type. Corresponding to each type we have a rough scale of credibility for different types of events. In cases where we know nothing of the witness beyond his bare statement that he witnessed the occurrence of the event our estimate of his credibility is based on very general and usually rather vague considerations. We are guided by two things: the initial credibility of the event and our estimate of the general value of human testimony. Both of these criteria, and in particular the second, are excessively ill-defined.