Errors of memory are much more difficult for a thoroughly honest person to avoid, since very few are aware of the untrustworthiness of their own memory. This ignorance aids unconscious tendencies to bring events into harmony with onés own beliefs and opinions - whether religious or scientific - and with each other; to exaggerate the clearness and precision of recollections, and to simplify them, either by bringing any group of events into a connected whole or merely by dropping some of the details. The total effect may be an exaggeration of the marvellous elements in a story, or occasionally the reverse; e.g. a waking hallucination may be remembered as a dream.

Having reviewed these various possible sources of error, we may now pass on to consider the canons of evidence in the case of telepathy. In a typical instance of a telepathic phenomenon, there are three points on which indisputable evidence is required: (1) that the agent has had an unusual experience, - say, has died; (2) that the percipient has had an unusual experience, including a certain impression of the agent, - say, has, while awake, had a vision of the agent in the room; (3) that the two events coincided1 in time, which implies that their respective dates can be accurately fixed.

(1) Of the agent's experience, evidence independent of the percipient's statements can generally be obtained, either from printed notices if he has died, or, if not, from the man himself. (2) The percipient's experience consists in an impression affecting himself alone, to which therefore no one else can bear direct witness. But if this impression is made known at once to some other person, so that the latter's confirmation may date from a time before the condition of the agent was known, such confirmation is valuable as being practically independent of the percipient. (3) Evidence is required that the experiences of the agent and percipient happened on the same day,2 since the closeness of the coincidence may easily be exaggerated in memory, especially after the lapse of some time.

The worth of the evidence of course varies greatly in different cases, and the evidential conditions may be summarised according to their value as follows: -

A. Where the event which befell the agent, with its date, is recorded in printed notices, or in contemporary documents which we have examined; or is reported to us by the agent himself independently, or by some independent witness or witnesses; and where:

(1) The percipient (a) made a written record of his experience, with its date, at the time of its occurrence, which record we have either seen or otherwise ascertained to be still in existence; or (ß) before the arrival of the news* mentioned his experience to one or more persons, by whom the fact that he so mentioned it is corroborated; or (y) immediately adopted a special course of action on the strength of his experience, as is proved by external evidence, documentary or personal.

1 Supposing the facts proved, we have still to meet the objection that the coincidence of events may have been due to chance and not to telepathy (see account of the Report on the Census of Hallucinations, 612 A, for discussion of this); but the first step obviously is to prove the facts.

2 Since numerical data are required to estimate the argument for chance coincidence, an arbitrary limit of time for the coincidence must be fixed, and we include all cases in which the interval between the two events is not more than twelve hours, though it appears generally to be much less than this.

3 The words "the news " mean always in this connection the news of what has befallen the supposed agent.

(2) The documentary evidence mentioned in (1α) and (1y) is alleged to have existed, but has not been accessible to our inspection; or the experience is alleged to have been mentioned as in(1ß), or the action taken on the strength of it to have been remarked as in (1y), but owing to death or other causes, the person or persons to whom the experience was mentioned, or by whom the action was remarked, can no longer corroborate the fact.

This second class of cases is placed here for convenience, but should probably rank below the next class. At the same time the fact that the percipient's experience was noted in writing by him, or was communicated to another person, or was acted on, before the arrival of the news, is not one which is at all specially likely to be unconsciously invented by him afterwards.

(3) The percipient did not (a) make any written record, nor (ß) make any verbal mention of his experience until after the arrival of the news, but then did one or both; of which fact we have confirmation.

This class is, of course, as a rule, decidedly inferior to the first class. At the same time, cases occur under it in which the news was so immediate that the fact of the coincidence could only be impugned by representing the whole story as an invention.

(4) The immediate record or mention on the arrival of the news is alleged to have been made, but owing to loss of papers, death of friends, or other causes, cannot be confirmed.

(5) The percipient alleges that he remarked the coincidence when he heard the news; but no record or mention of the circumstance was made until some time afterwards.

Such cases, of course, rapidly lose any value they may have as the time increases which separates the account from the incident. Still, sometimes we have been able to obtain the independent evidence of some one who heard an account previous to the present report to us; or we have ourselves obtained two reports separated by a considerable interval. And where a comparison of accounts given at different times shows that they do not vary, this is to some extent an indication of accuracy.

B. Where the percipient is our sole authority for the nature and date of the event which he alleges to have befallen the agent.

In many of these cases, the percipient is also our sole authority for his own experience; and the evidence under this head will then be weaker than in any of the above classes. But where we have independent testimony of the percipient's mention of the two events, and of their coincidence, soon after their occurence - he having been at the time in such circumstances that he would naturally know the nature and date of what had befallen the agent - the case may rank as higher in value than some of those of Class A (5).

The analysis just given refers exclusively to first-hand evidence, that is, evidence in which the main account comes to us direct from the percipient. There is one, and only one, sort of second-hand evidence which can on the whole be placed on a par with first-hand; namely, the evidence of a person who has been informed of the experience of the percipient while the latter was still unaware of the corresponding event; and who has had equal opportunities with the percipient for learning the truth of that event, and confirming the coincidence. The second-hand witness's testimony in such a case is quite as likely to be accurate as the percipient's, for though his impression of the actual details will no doubt be less vivid, yet on the other hand he will not be under the same temptation to exaggerate the force or strangeness of the impression in subsequent retrospection. The risks of error in all other second-hand evidence have been so abundantly proved by experience (some illustrations of this are given in this section of chapter iv (Sleep). of Phantasms, pp. 149 to 157) that it is better to leave it altogether out of consideration, and the great majority of the cases given in Phantasms are first-hand.1