But it is not the interval between event and publication which really matters much the important interval to abbreviate is that between occurrence and record. It does not really signify how long ago things happened, provided the record is contemporary. Professor Newcomb seems to mix up two distinct intervals of time, as others have done before him. Nevertheless, one disadvantage does attend even well-recorded incidents of long ago, namely, that many of the actors or witnesses must be dead, so that further and supplementary inquiry is hampered.

Unless events are recorded so as to be beyond the chance of invention, lapsed memory, and casual coincidence-to say nothing of deliberate sophistication-our aim is to exclude the narration of them altogether; and many a would-be contributor to our Journal has been dismayed by the stringency with which tests are applied and questions asked. A few weak cases must no doubt have been admitted, but extremely few, and never with our good will. Rather would we reject many sound cases than admit one feeble one. We do not wish to rely on weak evidence, or to present it even by way of illustration, still less as material from which any inference can be drawn.

So now we come to Professor Newcomb's second objection to group 2-that of chance. Are the veridical or coincidental casescorresponding in time with the death or other catastrophe betokened by them-more numerous and fuller of detail than can be accounted for by chance ? Or will chance coincidence furnish a normal explanation?

It is a question which has been under our consideration always, and from the first. The whole subject of coincidence and chance has received very careful attention at the hands of the Society, and 170 pages in Vol. XIV. of its Proceedings are occupied with an elaborate discussion of problems thus arising.

Without repeating anything that is there said, it is clear to common-sense that chance must be responsible for a greater crop of coincidences among a group of occurrences which are plentiful, than among those which are rare.

But surely, it will be said, dreams are extremely common, and some must therefore accidentally be fulfilled. Yes, they are, too commonnever evening wears to morning but most sleepers dream-and accordingly the Society has always admitted the much greater chance of casual coincidence in the case of veridical dreams.

But visions-actual hallucinatory figures or apparitions-and sounds, conveying impressions so clear and distinct as to be recorded and mentioned to others before actual correspondence is known : these are not very common among sane and healthy people. They seem not to be numbered even by hundreds per annum, certainly not by anything like millions. People liable to have them frequently are

encouraged by us to make a note of all such occurrences as they intend to count, whether they succeed or whether they fail. If they do not act on this suggestion, their record of successes has perforce to be ignored as inconclusive, for the data are incomplete.

But to a large number of percipients of this class the experience is unique in their lives-and in that case they are asked to testify to that effect. They do not quite see the bearing of this inquiry, and their natural tendency would be-assuming that they are given to exaggeration to claim for themselves something like a faculty for ghost-seeing. When they disclaim it, and are manifestly upset by the strangeness of the occurrence, they can be believed.

Nevertheless, the second objection-the plea of accidental coincidence, even of apparitions-must be faced. When shots are innumerable, some of them must hit. So if phantasmal appearances are really exceedingly numerous, if everybody has them, a large number must coincide with reality by sheer accident.

Well, now, this is an a priori possibility which in our Proceedings has been fully admitted, strongly emphasised, and definitely refuted. The census of hallucinations--a most laborious piece of workwas undertaken by Professor and Mrs. Sidgwick and others, entirely with the object of inquiring into the actual facts. Their aim was to find out what proportion of people do have definite hallucinatory visions, and not simply to assume, as Professor Newcomb does, that everybody has.

As many sane and healthy people as possible were to be asked, by a definite and considered form of question, whether they had had a single hallucination in their lives; and the statistical collectors were thoroughly instructed to regard the answer 'No' as just as valuable as the answer 'Yes.'

But Professor Newcomb urges that a certain number of coincidences must be due to chance. Granted. The only question is what number may be so expected. That is discussed in the Census Report, and to that I now turn.

It is clear that if all spontaneous hallucinations were veridical or coincidental, an explanation by chance would be absurd; but some of them are certainly not coincidental-they occur when nothing particular is happening to the person represented, so the Council of the Society realised strongly that an estimate must be formed of the proportion which one set of cases bears to the other.

Mr. Gurney was the first to begin such an inquiry, in the year 1885, and his results are given in Chapter XIII. of Phantasms of the Living; and his introductory pages, at the beginning of Volume II. of that work, are well worth reading. He obtained answers from 5705 persons, and, although this number is admittedly too small for safe deduction, yet, as far as they went, the results very strongly tended to negative the hypothesis of mere chance. He urged that a more

extended inquiry should be undertaken in due time. The matter was brought by Professor Sidgwick before the International Congress of Experimental Psychology, at its meeting in Paris in 1889, and, with the approval of that body, a special committee was appointed, with Professor Sidgwick as chairman, to carry out the statistical inquiry and to report. Of this committee the chief workers must have been Mrs. Sidgwick and Miss Johnson. The committee's Report, published in 1894, constitutes Vol. X. of the Proceedings of the Society for Psychical Research. It is not only well worthy of study, but its study is an absolutely essential equipment of anyone who at any future time attempts to discuss seriously the subject of the coincidence between phantasmal appearances and what they purport to represent.

Chapter XIII. of the Report is headed Chance Coincidence,' and the committee open it as follows:

We are now in a position to estimate the improbability that the death-coincidences are due to chance. The fact that each of us only dies once, enables us to calculate definitely the probability that that death will coincide with any other given event, such as the recognised apparition of the dying person.

Some figures are then quoted from the Registrar-General's Report for the decade 1881 to 1890, and, after discussion, the chance that any one person taken at random will die on any given day is estimated to be one in 19,000.

We ought, therefore, to find that out of 19,000 apparitions of living persons, or persons not more than twelve hours dead, one is a death-coincidence-occurs, that is, on the day of the death of the person seen, and within twelve hours of the death on either side.

Now of the 17,000 persons whose answers are included in the Report, those who had experienced hallucinations numbered 1684. Among these hallucinations the number of apparitions was 381; namely, 352 realistic apparitions, twenty cases of partial apparition, and nine visions of a person alive. But twenty-eight of the informants said that they had had several unreported hallucinations; and since the data are incomplete in their case, it was thought safer to omit this small group altogether from the numerical discussion.

The retained number of apparitions thus became 350 out of the 17,000 inquiries. An advocate of chance, however, would insist that these are only the apparitions remembered-that more may have been really seen and forgotten; thereby increasing the opportunity for casual coincidence with reality. So the probabilities of forgetfulness are carefully discussed in the Report of the Census Committee. Ultimately they decided to assume, as an extreme precaution, that perhaps three times as many hallucinations had been forgotten as remembered; thus raising the total number to a possible 1300, and allowing even exaggerated scope for the play of chance.

The next thing to ascertain was the number of death-coincidences

-of real and trustworthy death-coincidences-in this group; and here the path of safety lay not in increasing but in decreasing the number; so after making every allowance for possible exaggeration and selection, and excluding everything that could be considered in the least suspicious, they came to the conclusion that thirty safe death-coincidences were to be found among the 350 cases; that is to say, one in twelve about; or, making the above large allowance for forgetfulness, one in forty-seven. But this is equivalent to 400 in 19,000, or 400 times the most probable number.

Or, looking at the matter in a different way, if death-coincidences only occur by chance, 570,000 apparitions would be needed to produce thirty chance coincidences; and of the total number we may assume that about a quarter, or 142,500, would be remembered. That being so, we should expect to have to collect 142,000 cases, instead of only 350, in order to obtain thirty death-coincidences merely by chance.

But all this is based on the supposition that the apparition, in order to be counted as coincident, may follow or precede the death by as much as twelve hours on either side; whereas in the great majority of cases the coincidence in point of time is asserted to be far closer than this. And it is clear that if the apparition occurs within one hour of the decease, the probability against its chance occurrence is increased twelvefold.

The committee, therefore, conclude that the number of deathcoincidences in their collection is not due to chance; and they feel well assured that if anyone, with the most elementary acquaintance with the doctrine of chances, will critically examine their record, they will be bound to come to the same conclusion. Of that there is really no doubt, but I fear it is hardly to be hoped that opponents of a telepathic or other supernormal explanation will take this trouble. If they do, they must fall back on other lines of argument-such as misrepresentation, fraudulent collection, or some other device. That is legitimate, if they can substantiate such a claim, but the doctrine of chance coincidence is not legitimate: it is negatived in a scientific manner by the facts.

Assumption and prejudice, however, are powerful weapons in this subject-more powerful than calm and critical inquiry. It is easier and more effective to make plausible assumptions than laboriously to collect and discuss data.

An objection that may be made to the inquiry is that pathological phantasms are common enough; medical evidence is abundant for hallucination under the influence of drugs, or of illness, or of insanity. But none of these cases were included in the census; it was directed solely to the waking hallucination of sane and healthy people. And we find that such hallucinations are rare. Everyone may have momentarily mistaken an old coat or a shadow for a person; but that is an illusion, not a hallucination. An illusion is a wrong interpre

tation of an actual object. A hallucination is a perception as of an object which is not there; though in the veridical cases it is proved to correspond with some reality elsewhere, while in the non-veridical cases such correspondence is not established. Edmund Gurney's careful definition of a hallucination is the following:- A percept which lacks, but which can only by distinct reflection be recognised as lacking, the objective basis which it suggests.'

It will be asked, how do we know that pathological cases were excluded? How do we know that the instructions to collectors to avoid them were obeyed? Well, there is a definite answer to that, too. Since the census report, and independent of it, Dr. Henry Head published in a medical journal a report on hallucinations associated with visceral diseases, among which occur phantasms caused pathologically by diseases of the visceral system. These have certain generic characteristics, so that they constitute a distinct group.

In an S.P.R. paper (Proceedings, vol. xix. pp. 267-341), Mr. Piddington took the trouble to compare these pathological hallucinations with those recorded in the census; and thereby discovered that the census cases, i.e. the sane and healthy group, had totally different generic characteristics from the pathological group. Since that time we have felt even more confidence than before in the conclusion of our census committee.

But it may be further asked, What ground have we for attributing sane and healthy veridical phantasms to telepathic influence, at least as a working hypothesis?

One answer is that it is the least forced or supernormal explanation we can think of. But another answer depends on the following facts.

In addition to the spontaneous cases of phantasms, we have some experimental cases-that is to say, cases in which the percipient sees an apparition of someone who is trying to transfer an idea of himself to the percipient's mind, without any previous knowledge on the part of the latter that such an attempt is being made.

There are fifteen successful experiments of this kind already recorded by our Society, in which ten different experimenters have taken part. The records are all at first-hand, and in every case the evidence of the percipient has been obtained as well as that of the experimenter.

Nevertheless we do not, in any of these cases-whether spontaneous or experimental-make any positive assertion as to what the particular cause of the coincidental phantasmal appearance may be. More than one cause may exist, and different causes may operate in different cases. All we can say for certain is that in most cases the real and undoubted coincidence is not due to chance.

The final report of the committee is thus summed up:

Apparitions which coincide in time with the death of the person seen are the most important, because they are the most numerous, and because they afford