What, then, is the cause of the belief that a phantom of a man is a token of his death?  On the theory of savage philosophy, as explained by Mr. Tylor himself, a man’s soul may leave his body and become visible to others, not at death only, but on many other occasions, in dream, trance, lethargy.  All these are much more frequent conditions, in every man’s career, than the fact of dying.  Why, then, is the phantasm supposed by savages to announce death?  Is it because, in a sufficient ratio of cases to provoke remark, early man has found the appearance and the death to be ‘things connected in fact’?

I give an instance in which the philosophy of savages would lead them not to connect a phantasm of a living man with his death.

The Woi Worung, an Australian tribe, hold that ’the Murup [wraith] of an individual could be sent from him by magic, as, for instance, when a hunter incautiously went to sleep when out hunting.’[15] In this case the hunter is exposed to the magic of his enemies.  But the Murup, or detached soul, would be visible to people at a distance when its owner is only asleep—­according to the savage philosophy.  Why, then, when the wraith is seen, is the owner believed to be dying?  Are the things bound to be ‘connected in fact’?

As is well known, the Society for Psychical Research has attempted a little census, for the purpose of discovering whether hallucinations representing persons at a distance coincided, within twelve hours, with their deaths, in a larger ratio than the laws of chance allow as possible.  If it be so, the Maori might have some ground for his theory that such hallucinations betoken a decease.  I do not believe that any such census can enable us to reach an affirmative conclusion which science will accept.  In spite of all precautions taken, all warnings before, and ‘allowances’ made later, collectors of evidence will ‘select’ affirmative cases already known, or (which is equally fatal) will be suspected of doing so.  Again, illusions of memory, increasing the closeness of the coincidence, will come in—­or it will be easy to say that they came in.  ‘Allowances’ for them will not be accepted.

Once more, 17,000 cases, though a larger number than is usual in biological inquiries, are decidedly not enough for a popular argument on probabilities; a million, it will be said, would not be too many.  Finally, granting honesty, accurate memory, and non-selection (none of which will be granted by opponents), it is easy to say that odd things must occur, and that the large proportion of affirmative answers as to coincidental hallucinations is just a specimen of these odd things.

Other objections are put forward by teachers of popular science who have not examined—­or, having examined, misreport—­the results of the Census in detail.  I may give an example of their method.