§ 3. The points to be considered have to do both with the intensity and with the content of the dream; let us consider them in order.

First as to intensity. An exceedingly small proportion of dreams are remembered with distinctness several hours after waking. Even of the dreams which dwell in the memory, an exceedingly small proportion produce any appreciable amount of distress or excitement. And of these more impressive dreams, an exceedingly small proportion prove their intensity by being in any way acted on. What I have termed intensity may be indicated in another way, by the rapid repetition of a distinct dream two or three times on the same night; and this, too, when there is no apparent cause to prompt the dream, seems to be a comparatively rare occurrence. The dreams to be cited in this book will nearly all, I think, be distinguishable by one or other of these tests of exceptional intensity. And in proportion as the dreams which coincide with the event dreamed of are thus found to be in some other way exceptional - in proportion as the class to which they belong is found to form a small and sparse minority among the swarming multitude of unmarked dreams - in that proportion does it clearly become unreasonable to argue that the coincidences are sufficiently accounted for by the law of chances.

The argument which might seem effective so long as we had the whole multitude of dreams to range over - that multitude seeming sufficienttogive the law of chances ample scope - assumes quite a different aspect when we find ourselves limited to the comparatively small group of intense dreams.

Next as to content. Before we can give weight to a dream-coincidence as pointing to anything beyond the operation of chance, we should inquire whether the event dreamed of is distinct, unexpected, and unusual. If it combine all three characteristics in a high degree, its evidential value may be very considerable; in proportion as the degree falls short, or the combination fails, the evidential value sinks; and none of the characteristics taken alone, even though present in a high degree, would lead us to include a dream in the present collection. Thus, the dream-content must be neither a vague impression of calamity nor of happiness; nor a catastrophe on which the sleeper's mind is already fixed; nor some such ordinary event as has frequently occurred in waking experience. It may, indeed, be not the less significant for being trivial; but in that case it must be of a bizarre or unlikely kind. Then again, amount of detail, and the number of connected events, are of immense importance, as each subsequently verified detail tells with ever-mounting strength against the hypothesis of accidental coincidence. Once more, dream-content must be considered to some extent in relation to the dream-habits of the particular dreamer.

Before estimating the value of the fact that a person has dreamt of the sudden death of a friend on the night when the death took place, we should have to ascertain that that person is not in the habit of dreaming of distressing or horrible events. In respect of these various points, the instances to be cited, here and in the Supplement, are the sifted survival of many less definite coincidences in which the popular imagination would find a marvel. And in the case of this residue, where we have complete fulfilment of some of the above conditions, over and above the close proximity in time, or (it may be) absolute synchronism, of the event and the dream, the question as to a causal connection between the two is, at any rate, not to be swept out of court by a mere general appeal to the doctrine of chances.

But there is a further point in the content of the dreams that correspond with real events - true dreams, as we may for brevity call them - p which cannot but strike the attention as soon as we begin to examine actual specimens. It is that, among true dreams, by very far the largest class is the class where the truth is death - i.e., where the event dreamt of as happening to another is of that most restricted kind which can only happen once in each individual's experience. Out of the 149 coincident dreams which are included in this book - as at least clearly finding in telepathy, if it exists, their most natural explanation - no less than 79 have represented or suggested death. This, it will be seen, is entirely in accordance with a theory of causal connection between event and dream, where the abnormal state of the person dreamt of is regarded as part of the cause; but it is not at all in accordance with what we should expect accident to bring about. Nor could this argument well be met by the assumption that it is only in the case of a very grave event that the accidental correspondence attracts attention and gets recorded.

For this would mean that the dreams of death which happen to correspond with reality are one specially-remembered class among the total number of accidentally-true dreams. Now it will be admitted that dreams of death constitute a minute proportion of all dreams; it would follow, then, on the above assumption, that accidentally-true dreams of death constitute a similarly minute proportion of all accidentally-true dreams. But at this rate the total number of "true dreams" - in other words, the number of coincidences which the doctrine of chances will have to cover - swells to a most prodigious and unmanageable figure. It is just because a "true dream" is a very exceptional occurrence, that it was possible even to attempt to account for it as an accident; if the "accident" is for ever "coming off," so much the worse for that attempt.