§ 4. But the fact that a singular and marked event, such as death, is in so large a proportion of cases the central feature of the "true" dreams, supplies more than a general argument; it supplies the means for an actual numerical estimate whereby the adequacy of the chance-hypothesis may be tested. For dreams of so definite a character, and which admit of being so clearly and briefly described, are quite a fit subject for statistics; there is a possibility of finding out approximately what the actual frequency of a dream of this sort is; and we shall then have the first necessary datum for deciding whether the frequency of the cases where it coincides with reality, is, or is not, greater than chance would fairly allow. If it turned out that all of us about once a week had a marked and distressing dream of the death of some friend or other, then, since people who are somebody's friends are perpetually dying, the coincidence of such a dream with the real event might be expected to occur by pure accident in a large number of cases. But if only a small minority of us could recall ever having had such a dream at all, the case would be reversed.

The object, then, is to ascertain from a number of people, large enough and promiscuous enough to be accepted as a fair specimen of the whole population, what percentage of them have had the experience in question. With this view, efforts have been made, dating from the winter of 1883, to obtain a large number of answers to the following question: -

Since January 1st, 1874, have you ever had a dream of the death of some person known to you, which dream you marked as an exceptionally vivid one, and of which the distressing impression lasted for as long as an hour after you rose in the morning?

This question has been put to 5360 persons, as to whom it was not known beforehand whether their answer would be "Yes" or "No" Of these persons, 173 answered "Yes" Excluding 7 of these cases, in which the dreamer was at the time in a state of distinct anxiety as to the person whose death was dreamt of, we have a remainder of 166. These include a good many cases where it proved, on further inquiry, .that the terms of the question had not been exactly met, as the impression had not endured in any vivid or distressing way. They also include 3 cases where the mind of the dreamer had been exceptionally directed to the person dreamt of; and 3 cases where the person dreamt of was in the same room as the dreamer, which may have had some tendency to produce the dream - one gentleman expressly stating his suspicion that his dream was caused by some sound made by his companion. We may, therefore, accept the 166 as a liberal estimate.. But 18 of these persons professed to have had a dream of the sort inquired about more than once.

It will be again a liberal estimate if we suppose each of these to have experienced 3 distinct examples within the specified time; and the most convenient way of taking account of these repetitions will be to add 36 to the 166, making 202. With this substitution, 1/26 of the whole number of persons asked may be taken to have given an affirmative answer.1 Now, the persons asked were a quite promiscuous body, and a body large enough to be safely regarded as a fair average sample of the population; just as a similar number of persons, taken at random, would be accepted as a fair sample for purposes of statistics on short sight, or colour-blindness. We may conclude, then, that the number of persons who can recall having had - during the twelve years 1874-1885, and without special assignable cause - the experience named in the question, amount to about 1 in 26 of the population of this country.

The question next occurs, ought we, in making our calculation, to assign any limit to the area of acquaintances from whom the person dreamt of may be drawn, and of whom a certain proportion will, in the natural course of things, die within a period of 12 years? On general grounds it may fairly be argued that the slightness of connection between two persons would diminish the chance that one would dream accidentally, no less than the chance that he would dream telepathically, of the death of the other; and that therefore a vivid dream of the death of a stranger, or of a slight acquaintance, when it coincides with the death, tells neither more or less in favour of the action of something beyond chance than a similarly vivid and coincident dream of the death of a near relative. It will be seen, moreover, that, as far as the numerical estimate goes, it is unimportant how large or how small we take the area to be; because whatever number of persons we include, on the average the same proportion of the number will die within any given time.

Thus assuming this' proportion who die to be one-fourth of the whole number, then, if we took a very large circle of acquaintances, say 400 - the death of any of whom, if dreamt of when it occurred, would count as one of our coincidences - we should have to reckon that 100 of them actually die in the course of the time; and if we took a very small circle of just the immediate relatives of the dreamer, say 4, we should have to reckon that 1 of them dies in the course of the time. And the chance of an accidental coincidence in the specified period between the single dream of death and the death itself is practically the same in both these cases. For though there will be, on the one hand, much less chance of its being the right individual that is dreamt of when the choice is among 400 than when it is among 4, on the other hand the 100 deaths will give 100 (or nearly 100) nights on which the coincident dream will have its chance of being dreamt, instead of only 1 night. Logically, therefore, there is no necessity for limiting the area in question. Let the number of any one's acquaintance be called x.

Then, whether x be large or small, all that concerns us is the proportion of the x persons who die within the specified period of 12 years; and this proportion, since the death-rate per year is about .022, may be taken as .264, or a little more than one-fourth.

1 Some farther returns, received since this page was printed, have unaltered the proportion stated.

2 It is not always easy to find out, from a brief description, the strength of the bond that has existed between two persons. But I think I am safe in saying that the dreams of death in which the person dreamt of is not linked to the dreamer by a rather close tie of kinship or affection, do not amount to 10 per cent. of the whole number. In other respects the dreams do not seem to follow any line of a priori likelihood - e.g., they concern the death of young persons quite as much as of old ones.

I may point out that a different estimate would have to be made for dreams dreamt by several dreamers about the same person: e.g., an accidental coincidence of dream and death is less improbable than usual where the person dreamt of is a prominent public character, because he is (so to speak) within the dream-horizon of an immense number of people. But the proviso has no practical importance, as no cases of the kind occur in my statistics.