Charles McCreery

Charles McCreery

Graduate of the Oxford Honours Schools of Philosophy, Psychology and Physiology, and a Research Officer of the Institute of Psychophysical Research, Oxford. His books include "Science, Philosophy and ESP", "Psychical Phenomena and the Physical World" and "Apparitions" (with Celia Green).

Attempts at an Explanation of the Shackleton Experiments

 - Charles McCreery -

          THE FOLLOWING quotation is from a paper by Dr. Eric Berne entitled 'The Nature of Intuition'(1). The author is a psychiatrist.

(1) Berne, E., 'The Nature of Intuition', The Psychiatric Quarterly, Vol. XXIII, 1949, pp. 203-26.

At this point [in a professional interview] another man who was a stranger to me entered the office... He ... said nothing, but listened with great interest [while Dr. Berne continued to give his intuitive impressions about another man]... After listening to a few more such exchanges, the second man requested to be told something about himself.

Q. Well, I think your father was very strict with you. You had to help him on the farm. You never went fishing or hunting with him. You had to go on your own, with a bunch of rather tough fellows.

A. That's right.

Q. He began to scare you badly when you were about seven years old.

A. Well, my mother died when I was six, if that had anything to do with it.

Q. Were you pretty close to her?

A. I was.

Q. So her death left you more or less at the mercy of your father?

A. I guess it did.

Q. You make your wife angry.

A. I guess I did. We're divorced.

[This took me by surprise. After a moment we proceeded:]

Q. She was about sixteen and a half when you married her?

A. That's right.

Q. And you were about nineteen and a half when you married her?

A. That's right.

Q. Is it right within six months?

A. (Pause.) They're both right within two months.

Q. Well, fellows, that's as far as I can go.

A. Will you try to guess my age?

Q. I don't think I'm in the groove for guessing ages tonight. I think I'm through.

A. Well, try, sir.

Q. I don't think I'll get this but I'll try. You were 24 in September.

A. I was 30 in October.

Q. Well, there you are.

Some people may be inclined to think that ESP must have been at work in this situation. But despite the repeated accuracy with which, when in the mood, he finds he can guess at such details as a stranger's city of birth or the age at which he left home(2), Dr. Berne denies that his intuitions contain any ESP. He acknowledges their resemblance to ESP impressions, but believes they are derived from subliminal interpretation based on extensive past experience of facial expressions and other signs. He admits 'one is sometimes astonished at the accuracy of the intuition ... One would expect, if one guessed "number" in a large series of cases, to be right in a certain proportion; it is quite another thing to be right almost all the time when in a certain frame of mind'.

(2) Cf. Soal and Bateman: 'Shackleton claimed to be able to sum up a Person's character by a flash of intuition, and his thumb-nail sketches of strangers to whom he was introduced were often very amusing and accurate.'

Obviously one can never be certain that one's intuitions do not contain ESP; how would one set about trying to rule out the possibility? But at the same time how can one rule out the possibility that Dr. Berne's performance was all done on sensory cues?

In general, the senses (of certain people, at least) are some times capable of surprising feats of hyperacuity. In the 1880's, for instance, two Frenchmen, Bergson and Robinet, discovered a boy who, under hypnosis, appeared to be able to 'read' by clairvoyance book page numbers chosen at random by one of the experimenters(3). The conditions under which he could do this, however, were extremely specific: the experimenter had to stand with his back against the light, the book had to be held nearly vertical, about 4 in. from the experimenter's eyes, but slightly below them so that he could look sometimes at the page and sometimes at the boy. It occurred to one of the experimenters that the boy might conceivably be reading the numbers from their reflection in the cornea of the reader. F. W. H. Myers, in reporting this case, calculated that the corneal image would have been about 0.1 mm. in height, while two people who carried out some experiments on reading similar letters in this way found they could only succeed when the letters were about 10 mm. high. This may seem to put the boy's performance beyond the bounds of (normal) possibility, but there is no knowing to what extent he was endowed with abnormal visual acuity, and, furthermore, it is possible that, whatever the boy's normal acuity may have been, it was enhanced under hypnosis.

(3) This case is discussed by F. W. H. Myers in Human Personality, Longmans, Green and Co., London, 1903, Vol. 1, pp. 477-9.

However, for the reasons given in the previous chapter, no one has yet succeeded in showing how hyperaesthesia could account for Shackleton's scores in the situation described. At the moment at which Shackleton made his guess no single person had normal knowledge of the nature of the target card next to be turned up. The Agent did not know which it would be because, not having seen Dr. Soal's list of random numbers, he did not know which number would appear next at the hole in the screen; and Dr. Soal did not know which picture would be looked at next because he did not know in what order the pictures were placed in the Agent's box.

Could there have been some mistake - did Shackleton really get the scores that are claimed? It has been argued that a very small percentage of scoring errors would be sufficient to explain apparent deviations from chance in scores that are only marginally significant.

Empirically, there is little evidence that recording errors have ever been sufficiently numerous to account for an apparently significant card-guessing score. Professor Gardner Murphy once gave details of 175,000 vocal guesses which were recorded with only 175 mistakes - an average of only one mistake for every thousand trials(4). Anyone who has tried scoring runs of Zener cards will know that it is usually much easier to miss a hit than vice versa. As an illustration of this: Dr. J. A. Greenwood, in a re-check of 500,000 guesses, found that there had been only ninety mistakes, out of which seventy-six were hits that had been missed(5).

(4) Murphy, G., 'On Limits of Recording Errors', Journal of Parapsychology, Vol. II, 1938, pp. 262-6.
(5) Greenwood, J. A., 'Analysis of a Large Chance Control Series of ESP Data', Journal of Parapsychology, Vol. II, 1938, pp. 138-44.

It is clear that errors of the frequency observed in these two examples could never account for scores of the size of Basil Shackleton's. One would have to suppose that the experimenters had made an average of two (positive) recording errors per run of twenty-five guesses! But a record of Shackleton's every guess is to be seen in the archives of the Society for Psychical Research. It is hardly likely that errors of this frequency would have gone undetected for twenty years. Besides, at the time of the experiments the original scoring sheets were checked by eight other people besides the experimenters and 'in all the original scoring sheets so checked over the total period of the experiments fewer than a dozen isolated errors were found, none of these being in the precognitive groups' (i.e. the groups in which Shackleton obtained his significant scores).

An alternative to the Recording Error Theory is the Theory of Unfair Selection. One of the objections brought against the first significant experiments reported by Rhine, for instance, was that they represented a selection from a far greater number of insignificant results that never found their way into print. (Clearly, if only one experiment in 100 gives odds of 100 to 1, it would hardly be fair to publish every 100th experiment as evidence for ESP.)

However, this theory failed to explain the results obtained by Rhine and it likewise fails to explain the results of Soal with Shackleton - for the following reason. Considering only the results already quoted - the score of 1,101 correct guesses out of a possible total of 3,789 between 24th January 1941 and 21st December 1941 - the probability of this result having occurred by chance is less than 10-35, i.e. 1 divided by a figure with thirty-five noughts in it. But it has been calculated that, even if the Shackleton series were the only significant ESP experiment that had ever been conducted, and even if 'every inhabitant of the globe had done an unsuccessful parapsychological experiment every month since the beginning of the tertiary period sixty million years ago',(6) the probability of Soal's results with Shackleton being simply due to chance would still be 10-17.

(6) Thouless, R. H., 'Thought Transference and Related Phenomena', Discourse to the Royal Institution, 1st December 1950.

It is sometimes argued by writers on the theory of probability that a statement of odds of the order of magnitude of 1035 is meaningless because it cannot possibly be verified. It is true that it would be quite impracticable for anyone to start matching random numbers 'by hand' with a view to seeing if the number of hits ever reached a level of significance comparable to that achieved by the card-guessing scores of Shackleton. However, this is only an empirical difficulty and does not amount to unverifiability in principle. It is possible that one day a computer will be built that is capable of generating as many random numbers in a few minutes as the whole population of the earth could have generated if it had been tossing coins since the beginning of the tertiary period. Thus to say that odds of 1035 to 1 are 'meaninglessly large', so far from constituting an objection to Soal's experiments, amounts to an admission that the 'unfair selection' theory will never be able to explain his results.

A more sophisticated possibility is that the random number sequences used by Soal were not completely random and the resulting non-randomness in the target cards happened to 'fit' with Shackleton's guessing habits in a way that produced spuriously extra-chance scores. To take a simple example: suppose Shackleton had a preference for calling 'Zebra'; and suppose there happened to be more Zebras in the target series than would be predicted by probability theory; then would not Shackleton have made more than five correct guesses per run of twenty-five without ever doing ESP?

Such a suggestion appears to gain some support from the fact that when random numbers have simply been matched against other random numbers the number of 'hits' has sometimes been found to differ significantly from the number probability theory would predict.

However, it is not enough to show that the published random number tables are biased; one has got to show that such a bias is responsible for the high scores achieved by Shackleton, and this has never been done.

It might, of course, be argued that it would be of some significance even if one could show that scores comparable to those obtained in the best card-guessing experiments can be obtained by matching runs of random numbers. It would then be fair to conclude that such scores as Shackleton's can be a statistical artifact, even if it remained only an inference that they were. But, again, this has not yet been done.

Moreover, it is not only the magnitude, it is the nature of Soal's results that are inexplicable on the assumption of target bias. For instance, it is not clear how any of the remarkable results described in the preceding chapter could be due to nonrandom tendencies in the random numbers used.

Of course, if anyone did succeed in obtaining odds in a 'pure probability' experiment that were comparable to those obtained in the Shackleton series, he would not be content to conclude merely that such things were possible in parapsychological experiments generally; he would turn to the Shackleton records and see if his discovery applied to them in particular. But quite apart from any reasons which might lead one to conclude that no one ever will make such a discovery, there is reason for supposing it would never apply to the Shackleton results. This reason is as follows. The previous chapter mentioned as being 'in many ways the crucial feature' of Soal's experimental procedure the fact that the Agent had only five target cards and that he or she re-shuffled these after each run of fifty guesses. Now, even if there did at one point exist a correspondence between Shackleton's guessing preferences and a non-random peculiarity of Soal's prepared fists of numbers, this correspondence would not have survived the scrambling of the actual target cards. Suppose, for instance, as we did above, that Shackleton had a preference for calling 'Zebra'; and suppose that there happened to be a predominance of threes in Soal's random number sequences; finally, imagine a run of guesses during which the 'Zebra' card occupied the third place from the left in the box. Under these conditions Shackleton could have got more than five guesses right without doing ESP. But the odds are four to one against 'Zebra' being in the third position during the next run of fifty guesses, so Shackleton's advantage would be short-lived.

Of course, it is possible that Shackleton's guessing-preferences kept changing and in this way continued to correspond to the peculiarities in the random number sequence. But this amounts to saying that he was using ESP to discover what changes were taking place in the order of the cards in the box!

We may conclude that no particular non-randomness in Soal's number sequences could have accounted for Shackleton's results. But perhaps 'probability theory' is fundamentally at fault when it tells us that there is something 'significant' in Shackleton's results in the first place?

If this were the case, we could, of course, jettison probability theory. But if we did, we should have to jettison most of modern science with it! Statistics are now used in nearly every branch of science, from quantum mechanics to radio-astronomy.

Moreover, there is at least one good reason for not getting rid of probability theory and that is that it works. There is nothing in the least mystical about this fact. (One of the first applications of Gauss and Laplace's 'Normal Law of Error' was to the girth of Scottish soldiers.) Consider the origins of probability theory. The word 'chance' comes from the Latin 'cadentia', a word used to describe the fall of dice; and what has been called 'the first contribution of mathematics to the theory of probability'(7) was made in 1654 by Pascal, in connection with a problem set him by his friend the Chevalier de Méré - a problem that had arisen in the course of the latter's gambling activities. The question set to Pascal was this: what is the probability of getting at least one double-six in the course of twenty-four throws with a pair of dice? The Chevalier argued that it must be the same as the probability of getting one six in four throws of a single dice; but in this he was wrong, as he discovered to his cost.

(7) Kneale, W., Probability and Induction, Oxford University Press, 1949, p. 123.

This example illustrates the fact that the foundations of so-called 'probability theory' are essentially empirical. All one needs for the solution of the Chevalier's problem is a certain empirical datum - the fact that if one makes a series of throws with an unbiased dice the proportion of sixes to other faces that most frequently occurs is one in six - and the rules of arithmetic with which to manipulate such data. The whole of the 'Calculus of Chances' is built up by means that are no more mysterious than this. It is open to anyone to question why the probability of throwing a six is what it is, and it is even open to anyone to question the validity of the rules of arithmetic. But granted these two things, there can be no cavilling at the uses to which statisticians put them.

It may be objected that there is no a priori reason why probability laws should describe the universe. But as it happens they do; and if they break down in one place and one place only (card-guessing experiments), then we must look for an explanation.

Besides, even if it can be shown that there is bias in the published random number tables, it is not clear why we should therefore reject probability theory - and with it the hypothesis of ESP. Why not reject the random number tables? If the published tables are not random then the answer is to construct better ones. (Though, the published tables being quite good enough for most purposes, one should rather say 'new ones for certain specific purposes'.)

To conclude: It is logically permissible to decide not to reject the null hypothesis even in relation to Shackleton's scores. However, experiments like Dr. Soal's are not subject to the first of the two limitations on spontaneous cases mentioned at the end of Chapter I. That is to say, we can assign a definite numerical value to the probability of the overall scores having occurred by chance. In the case of the nineteen sessions between 24th January 1941 and 21st December 1941, the frequency with which the overall score would recur by chance in an indefinite number of repetitions of the entire series of nineteen sittings is once every 1035 times.

To give some idea of the size of this figure: let the line A B in the figure below represent a length of 1035 units, then the line C D represents a length of 1034 units. Alternatively, if the line A B be taken to represent a length of 1034 units then the line C D represents a length of 1033 units. By extrapolation, the lines A B and C D may be taken to represent any two similarly related orders of magnitude.

In fact, if the line C D (which is about 10 min. long) was itself multiplied by a factor of 1035, the result would be a line a hundred thousand million million light-years long.

A __________________________________B


If one decides not to reject the null hypothesis one also has to maintain that it was 'just coincidence' that Shackleton got only chance scores with certain agents and extra-chance scores with others. In fact one would have to argue similarly concerning all the scoring-phenomena described in Chapter II (and a large number of others which are described in Soal's account but not reproduced in this book).

Alternatively, one can maintain that Dr. Soal and an indeterminate number of other people combined in perpetrating an elaborate hoax extending over a period of several years. However, if one asserts this view in the absence of any evidence to support it, and simply because one is unwilling to admit that something occurred for which we at present have no explanation, then one's assertion thereby invalidates the whole of science. To refuse to consider Soal's experiments as evidence for ESP just because they do not rule out the possibility of wholesale fraud is the same thing as to refuse to consider any experiment at all as evidence for ESP - since all evidence ultimately depends on human testimony. It is of course logically permissible to assert that the whole world is in a conspiracy against one, but to assert this is to deny the possibility of science.

Finally, one might conclude that phenomena occurred for which we at present have no explanation; with the corollary that if any explanation for the Shackleton experiments is ever discovered, this may yet be an explanation in terms of some already known hypothesis.

In Chapter X ('The Psychology of ESP') we shall be making the hypothesis that 'there is no qualitative difference between the state of mind in which ESP occurs and that in which PK occurs'. In the three chapters that follow we shall examine the ostensible PK phenomena produced by two outstanding subjects, Rudi Schneider and Eusapia Palladino, and consider whether these are explicable on any known hypothesis.


The article above was taken from Charles McCreery's "Science, Philosophy and ESP" (1967, Faber & Faber Ltd).

Related Material

Home | About Us | Latest News | Investigators | Articles | Experiments | Photographs | Theory | Online Library | Links | Recommended Books | Contact Us | Glossary | Search


Some parts The International Survivalist Society 2004