As an instance of the gradual approximation of dates, I may cite a case recorded in the _Proceedings of the American S.P.R._ (pp. 401, 527). The narrator wrote to Dr. Hodgson:--"I once dreamed that W. T. H. was dead; and the same night he was thrown down several feet on to an engine, ... when he was taken up it was thought he was dead." From later inquiries it was ascertained that the accident did indeed occur as alleged--but a week or ten days after the dream![77] As an illustration of a different kind of metamorphosis, a case may be given which I recently received from a lady and her daughter--an account of a "ghost" seen twenty-five years ago by the latter and her nurse. The younger lady described to me the figure seen; the mother told me that she had received a similar description from both nurse and daughter at the time of the incident. Both ladies were clear-headed and sensible witnesses, and it was impossible to doubt that they believed what they said. But in her childish diary, which the younger lady kindly unearthed for my inspection, the only entry referring to the matter--an entry written in pencil and obviously as an afterthought--ran: "Ellen saw a ghost." If the diarist had herself shared the experience, it is difficult to believe that even the modesty natural to her age and sex would have withheld her from recording the fact for her private glorification. It would be easy to multiply cases of this kind. But those who demand most proof of the action of telepathy will probably be least exacting of evidence for the untrustworthiness of ancient memories. As a matter of fact, we have the evidence of statistics to show that the imagination does tend after a certain lapse of time to magnify coincidences in matters of this kind, and even to invent coincidences where none existed. It will be shown in Chapter IX., in the discussion on the results obtained from an inquiry into the distribution of sensory hallucinations, that whereas non-coincidental hallucinations tend to be forgotten after the passing of a few years, the records of coincidental hallucinations--or at least of those which are alleged to have coincided with the death of the person seen--are proportionately more frequent ten years ago than at the present time, the inference being that a certain number of coincidences have been unconsciously improved or invented in the interval. _Pseudo-presentiment._