During the ensuing year, the Committee, consisting of Professor Barrett, Mr. Myers, and the present writer, made a number of experiments under similar conditions, which excluded contact and movement, and which confined the knowledge of the selected object - and, therefore, the chance of collusion with the percipient - to their own group. In some of these trials, conducted at Cambridge, Mrs. F. W. H. Myers and Miss Mason also took part. In a long series conducted at Dublin, Professor Barrett was alone with the percipient. Altogether these scrupulously guarded trials amounted to 497; and of this number 95 were completely successful at the first guess, and 45 at the second. The results may be clearer if arranged in a tabular form.

Table Showing The Success Obtained When The Selected Object Was Known To One Or More Of The Investigating Committee Only

1A full pack was used, from which a card was in each case drawn at random.

2This number is obtained by multiplying each figure of the third column by the corresponding figure in the fourth column (e.g. 216 x 1/52), and adding the products.

3This entry is calculated from the first three totals in the last horizontal row, in the same way that each other entry in the last column is calculated from the first three totals in the corresponding horizontal row.

Mr. F. Y. Edgeworth, to whom these results were submitted, and who calculated the final column of the Table, has kindly appended the following remarks:-

"These observations constitute a chain or rather coil of evidence, which at first-sight and upon a general view is seen to be very strong, but of which the full strength cannot be appreciated until the concatenation of the parts is considered.

"Viewed as a whole the Table presents the following data. There are in all 497 trials. Out of these there are 95 successes at the first guess. The number of successes most probable on the hypothesis of mere chance is 27. The problem is one of the class which I have discussed in the Proceedings of the S.P.R., Vol. III., p. 190, etc. The approximative formula there given is not well suited to the present case,1 in which the number of successes is very great, the probability of their being due to mere chance very small, in relation to the total number of trials. It is better to proceed directly according to the method employed in the paper referred to (p. 198) for the appreciation of M. Richet's result EPJYEIOD [see below, p. 60]. By this method,2 with the aid of appropriate tables,3 I find for the probability that the observed total of successes have resulted from some other agency than pure chance '999, 999, 999, 999, 999, 999, 999, 999, 98 " Stupendous as is this probability it falls short of that which the complete solution of our problem yields.

For, measuring and joining all the links of evidence according to the methods described in the paper referred to, I obtain a row of thirty-four nines following a decimal point.

A fortiori, if we take account of the second guesses.

"These figures more impressively than any words proclaim the certainty that the recorded observations must have resulted either from collusion on the part of those concerned (the hypothesis of illusion being excluded by the simplicity of the experiments), or from thought-transference of the sort which the investigators vindicate".

A large number of trials were also made in which the group of agents included one or more of the Creery family; and as bearing on the hypothesis of an ingenious family trick, it is worth noting that - except where Mr. Creery himself was thus included - the percentage of successes was, as a rule, not appreciably higher under these conditions than when the Committee alone were in the secret. When Mr. Creery was among the agents, the average of success was far higher; but his position in the affair was precisely the same as our own; and the most remarkable results were obtained while he was himself still in a state of doubt as to the genuineness of the phenomena which he was investigating.

One further evidential point should be noted. Supposing such a thing as a genuine faculty of thought-transference to exist, and to be capable, for example, of evoking in one mind the idea of a card on which other minds are concentrated, we might naturally expect that the card-pictures conveyed to the percipient would present various degrees of distinctness, and that there would be a considerable number of approximate guesses, as they might be given by a person who was allowed one fleeting glimpse at a card in an imperfect light. Such a person might often fail to name the card correctly, but his failures would be apt to be far more nearly right than those of another person who was simply guessing without any sort of guidance. This expectation was abundantly confirmed in our experiments. Thus, in a series of 32 trials, where only 5 first guesses were completely right, the suit was 14 times running named correctly on the first trial, and reiterated on the second. Knave was very frequently guessed as King, and vice versa, the suit being given correctly. The number of pips named was in many cases only one off the right number, this sort of failure being specially frequent when the number was over six.