But here the reader may fairly ask where the line of error is to be drawn. Must the coincidence be exact to the moment? And, if not, what degree of inexactness may be permitted before we cease to regard a case as supporting the telepathic hypothesis? It is unfortunately not easy for the moment to give any satisfactory answer to this question. Two distinct questions are in fact involved. The first is a question of natural fact: What are the furthest limits of time within which it appears, on a review of the whole subject, that a single telepathic phenomenon may really be included? At what distance of time, from the death of an absent person, may a friend receive telepathic intimation of the fact? The second question is one of interpretation and argument. It will be a most important part of our task hereafter to estimate the probability that it was by chance, and not as cause and effect, that the two events occurred at no very great distance of time from one another. The wider the interval, the greater, of course, does this probability become; in other words, the larger the scope that we give to "coincidences" which we are willing to regard as prima facie telepathic in origin, the greater is the chance that we shall be wrong in so regarding them.

Now, unless some provisional limit were assigned to the interval which may separate the two events, it would be impossible to obtain numerical data for calculating what the force of the argument for chance really is, and how far the hypothesis of some cause beyond chance is justified. This point will be made clear in Chapter XIII (The Theory Of Chance-Coincidence)., which deals with "the theory of chance-coincidence"; meanwhile it will be convenient to defer both these questions, and to make the following brief statement without discussion or explanation.

There is one class of cases which are not available for a numerical estimate at all - those, namely, where the agent's condition is not strictly limited in time; for instance, where he is merely very ill, and no particular crisis takes place at or near the time when the percipient's impression occurs. This indefiniteness is, of course, a serious evidential weakness. But in a vast majority of the cases to be brought forward, the event that befalls the agent is short and definite. If, then, the experience of the percipient does not exactly coincide with that event, it must either follow or precede it. And, first, if it follows it; then it will be convenient to limit the interval within which this must happen to 12 hours. I may mention at once that in most of our eases the coincidence seems to have been very considerably closer than this. But in a few cases the 12 hours' limit has been reached; and if we found that, though some error in evidence had made the coincidence appear to have been "closer than it really was, yet after correction the 12 hours seemed not to have been over-passed, we should still treat the case as having a prima facie claim to be considered telepathic.

Next as to the cases where the percipient's impression precedes some marked event or crisis in the existence of the other person concerned; the question will then be, What was that other person's condition at the actual time that the impression occurred? If it was normal, we should not argue here for any connection between the experiences of the two parties. For instance, we should not treat as evidence for telepathy an impression, however striking, which preceded by an appreciable interval an accident or sudden catastrophe of any sort. But it may happen that the percipient's impression falls within a season in which the condition of the other party is distinctly [abnormal - say a season of serious illness; and that it likewise precedes by less than 12 hours the crisis - usually death - with which that season closes. And these cases will not only have a prima facie claim to be considered telepathic, but will also admit of being used in a strict numerical estimate.

§ 11. To return now to the evidential question, it is really in the matter of dates, rather than facts, that the risk of an important mistake is greatest. In the first place, dates are hard things to remember: many persons who have a fairly accurate memory for facts which interest them have a poor memory for dates. This is a natural failing, and it is also one that may easily escape notice; for in the vast majority of instances where a personal experience is afterwards recounted, the whole interest centres in the fact, and none at all in the date. But in examining the evidence for an alleged telepathic case, much more than ordinary human frailty in the matter of dates has to be considered. It is just here that the action of the various positive tendencies to error, above enumerated, is really most to be apprehended. Two unusual events - say the death of a friend at a distance, and the hearing of a voice which certainly sounded like his - have happened at no very great distance of time. The latter event recalls the former to the mind of the person who experienced it; and on reflection he feels that the character of the one connects it in a certain way with the other.

True, he has kept no record of the day and hour when he heard the voice; or his friend may have died in South America, and no accurate report of the date of the occurrence may ever have reached England; but the connection which has been surmised cannot but raise a presumption that the two events corresponded in time as well as in character. "Why, otherwise, should I have heard the voice at all?" the person who heard it will argue: "I am not given to hearing phantasmal voices. I did not know how to account for it before; but now I see my way to doing so." This train of thought being pursued, it will seem in a very short time that the two events must have been simultaneous; and what can that mean but that they were simultaneous? And the fact thus arrived at will remain the point of the story, as long as it continues to be told. In allowing his mind to act thus, it will be seen that the percipient has merely followed the easy and convenient course. There was something baffling and aimless in the occurrence of the phantasmal voice, without rhyme or reason, at a time when the hearer was in good health and not even thinking of his friend. Rhyme and reason - significance and coherence - are supplied by the hypothesis that his friend, finding death imminent, was thinking of him.