Yet one more point must be touched on, to avoid misconception of the phrase cited above, that "the moment of death is the centre of a cluster of abnormal experiences, of which some precede, while others follow, the death." Gurney, of course, did not mean to assume that the act of death itself was the cause of all these experiences. Those which occur before death may be caused or conditioned, not by the death itself, but by the abnormal state, as of coma, delirium, etc., which preceded the death. This we say because we have many instances where veridical phantasms have coincided with moments of _crisis_--carriage-accidents and the like--occurring to distant agents, but not followed by death. Accordingly we find that in almost all cases where a phantasm, apparently veridical, has _preceded_ the agent's death, that death was the result of disease and not of accident. To this rule there are very few exceptions. There is a case given in _Phantasms of the Living_ (vol. ii. p. 52), where the phantasm seems on the evidence to have preceded by about half an hour (longitude allowed for) a sudden death by drowning. In this case the percipient was in a Norfolk farmhouse, the drowning man--or agent--was in a storm off the island of Tristan d'Acunha; and we have suggested that an error of clocks or of observation may account for the discrepancy. In another case the death was in a sense a violent one, for it was a suicide; but the morbidly excited state of the girl a few hours before death--when her phantasm was seen--was in itself a state of crisis. But there are also a few recorded cases (none of which were cited in _Phantasms of the Living_) where a phantasm or double of some person has been observed some days previous to that person's accidental death. The evidence obtained in the Census of Hallucinations, however, tended to show that cases of this sort are too few to suggest even _primâ facie_ a causal connection between the death and the apparition (see _Proceedings_ S.P.R. vol. x. p. 331). I now proceed briefly to review some of the cases where the interval between death and phantasm has been measurable by minutes or hours. It is not easy to get definite cases where the interval has been measurable by _minutes_; for if the percipient is at a distance from the agent we can seldom be sure that the clocks at both places have been correct, and correctly observed; while if he is _present_ with the agent we can rarely be sure that the phantasm observed is more than a mere subjective hallucination. Thus we have several accounts of a rushing sound heard by the watcher of a dying man just after his apparent death, or of some kind of luminosity observed near his person; but this is just the moment when we may suppose some subjective hallucination likely to occur, and if one person's senses alone are affected we cannot allow much evidential weight to the occurrence.[137]