TELEPATHY TESTS

Evening Star, Issue 22913, 22 March 1938, Page 14

TELEPATHY TESTS

RHINE EXPERIMENT STATISTICIANS APPROVE IT CHANCE EXCLUDED University psychologists are busily engaged in repeating the experiments of Professor J. B. Rhine, of Duke University—experiments which have convinced him that telepathy and clairvoyance are realities, and not the delusions of wishful thinking (writes the science editor of the ' New; York Times ’). Some of the professors have frankly set out to disprove “ extra sensory perception,” as Rhine calls it; others are seriously trying to settle the issues raised as objectively as possible. Whether Rhine is right or wrong depends at this largely on his mathematical case. Strange as it may seem, nearly all the statisticians of note find no fault with his conclusions', though they criticise his mathematical methods. It is the psychologists who object. Rhine uses a pack of cards consisting of five identical suits of five cards each. The cards of a suit bear simple geometrical designs—a circle, a star, parallel wavy lines, a cross or plus sign, a rectangle. After the pack has been shuffled it is laid face down and the percipient is asked to call the top card. His call is noted, whereupon the card is laid aside. So with the next card and the next, until the whole pack has been called. COMPARISON WITH CHANCE. If chance alone is involved, the percipient ought to guess correctly, on the average in a long series, one card in five or five in 25. But percipients gifted with E.S.P., _ as Ruin? calls extra sensory perception for short, cal! six, seven, oven nine, day in and day out for months. He concludes on the basis of trials which now number several hundred thousand that chance hits' cannot explain these results. _ The psychologists—not the statisticians—argue that when the -top_ card is laid aside only 24 cards remain, so that the chance of making a hit is no longer what it was. This would be good logic if the subject were told whether or not his first call were correct.

Suppose, for example, that the percipient has called the first card a star and is told that his guess is correct. Then he knows that there remain only four stars among the 24 untouched cards, so the chance that the next card will also be a star is only 4-24, which is less than 1-5. Similarly, if he has called the first card a star and is told that he is wrong, he knows that _ there are five stars among the remaining 24 cards, so the chance that the next card will also be a star is 5-24, which is greater than 1-5. Naturally, he will be influenced by such knowledge if he has it. But in (Rhine’s exiperiments the percipient is not told whether any call is right or wrong. If we assume that chance alone is responsible for his successes, his first attempt will fail four times as often as it succeeds. Hence for every successful occasion, with its chance of 4-24 that a star will appear on the second trial, there are four unsuccessful ones with chances of 5-24. CHANCES UNAFFECTED. In five such trials the percipient would .expect just 4-24 plus four times 5-24 stars to appear, and this works out at exactly one. In other words, if he does not know whether his first chance was correct or not the chance that the second card is a star is still just one in five. Another criticism which the psychologists have advanced stands on somewhat firmer ground in the sense that it is logically sound, but the statisticians dismiss it at once as a triviality. According to this criticism, Rhine has used a “ normal ” distribution curve in making his computations, whereas the correct distribution curve is certainly not “ normal.” To understand this argument we must first know what a distribution curve is.

If a subject calls just 25 cards, and if his successes are due to chance alone, we would expect him to get five right on the average. This does not mean that he would get exactly five right in each sequence of 25, Sometimes it would be six or seven, or three or four. More rarely one or two, or eight or nine. Very seldom outside these limits. If we construct a curve which snows how often each of these results can be expected to happen, it is called a “ distribution curve.”

Nobody knows what the correct distribution curve for the ESP experiments is. Even the distribution curve for the somewhat simpler problem of matching one shuffled pack against another has not been worked out in detail, though some recent work reported in the weekly ‘ Science ’ by Professor E. V. Huntington, of Harvard University, and Dr T. E. Sterne, of Harvard College Observatory, shows that it is very nearly normal. To do more than Huntington and Sterne have done even with this purely mechanical problem would acquire very laborious mathematics. CANNOT BE DETERMINED. * In the case of Rhine’s experiments, moreover, the difficulties are not entirely mathematical, for the correct distribution curve depends somewhat upon the mental habits of the percipient and will be a bit different if a subject shows a strong tendency to call circles much more frequently or plus signs much less often than he ought to. Such tendencies would not change the average number of successes, which would still be five, even in the extreme case of a subject who called every card a circle. They would, however, after the frequency of six or seven, or eight or nine hits. Because of these psychological factors, we can safely predict that no one will ever know just what the correct distribution curve for the Rhine experiments is, just as no one will ever know the correct curve for most of the statistical experiments that are carried out in biology or educational psychology. This is not as serious as it at first appears, for it is known _ that in all such experiments the distribution function approaches closer and closer to a certain special type as the number of trials is made larger and larger. This “ normal ” type, which was first studied by the great mathematician Gauss in the early nineteenth century and has been the standard for almost all statistical work since that time, is the one which Rhine uses. From it he finds that, if his data are the result of chance alone, the odds against calling 74 out of 74 cards correctly, which is what two of his percipients once did. are 1,000,000,000.000,000,000,000,000,000 to 1. It is these odds that the psychologists contend, and the statisticians-ad-mit. are wrong. Perhaps, say the statisticians, there ought to be three or more zeros, or perhaps three or four less, but since nobody has any con-

ception of the magnitude of the number represented by a one followed by 27 zeros, it doesn’t really matter much if we change the number of zeros to 23 or 31. All that matters is that the odds are stupendous in any case. NOT OF IMPORTANCE. In other words, for the practical purposes of life it does not, matter lion by another. So eminent a mathecurring is one in ten billion by one method (Rhine’s) or one in nine billion by another. So emient a mathematician as 11. A. Fisher, of London, expresses' the opinion that there is nothing radically wrong with Rhine’s conclusion that chance is not at work. Statisticians would rather have seen Rhine use the “ chi-square method ’’ of the late Professor Karl Pearson. Two of Rhine’s statisticians, Stuart and Greenwood, did use it m conjunction with the _ binomial distribution, but only sometimes. The chi-square method would have made it possible to determine the “ goodness of fit ” of any apparently suitable curve. On the other hand, it is generally conceded that oven though the normal probability curve does not fit any of chance occurrences well enough, Rhine’s general conclusions are not open to doubt. Dr Chester Kellogg, professor of psychology in M'Gill University, and on© of Rhine’s ■ most determined opponents, has written that once. he held a hand at bridge which consisted of all the spades in the pack, “ an event to be expected only onoe in 779,737,580,160 times,” and argues that “ as the same is true of every other hand . . . there was really no occasion for great wonder.” This hardly fits some of Rhine’s eases. Suppose the dealer had announced “ I am now going to deal you a hand of spades alone,” and suppose that he had actually dealt such a hand. If he succeeded, even Professor Kellogg would gasp just a little. So when two of. Rhine’s subjects manage to call 25 cards out of 25 correctly on isolated occasions, it may have been chance. But it may also be a link in a long chain of evidence in favour of extra sensory perception. SELECTION CRITICISED. Professor Kellogg makes the point that Rhine’s performers managed to guess more cards correctly in the first five and last five of a pack. This is sometimes true, hut not generally .true. Some of Rhine’s subjects guessed the middle groups of five better than the end groups. And the Juts for the middle groups were always more numerous than the expected score demanded. ... Rhine has been sharply criticised because he carefully selected his subjects. Ho mad© a series of preliminary

tests to weed out the scorers who did no better than they should statistically. It is objected that this is assuming in advance what is to be proved. Suppose we set out to discover whether there is such a phenomenon as skill in sharpshooting. If we test 100,000 men on the rifle range it is not likely that we shall reach sound conclusions on skill. Low scores would cancel high scores, and the statistical result would have no significance beyond showing what chances 100,000 men have of making a bull’s-eye. In other words, our conclusions would hold good only for the average rifleman. So far as the curve of distribution of hits shows, the very high scores may be chance successes. . . Extra sensory perception is like sharpshooting, if Rhine is right. Averago men and women are no more capable of calling cards correctly than they are of making bull's-eyes with a rifle time and time again. If extra sen- ( sory perception is to be studied it must • be with the aid of those who seem to logic is that followed in the physical sciences. To account for the perturbations in the orbit of Uranus the existence of an unknown outer planet was assumed. Later Neptune was discovered and the assumption confirmed. Then it was found necessary to assume that there must be a transNeptunian body. Again presupposition was justified*, this time by the discovery of Pluto, So in atomic physics, where such particles as electrons, protons, neutrons, were, assumed to exist long before their tracks were photographed. ASSUMPTIONS NEEDED. It is impossible to advance in science without making these assumptions. But verification is always necessary. In psychology the statistical verification of the assumption that there is extrasensory perception is dismissed as a piece of self-deception; in physics it would be proof of a theory. Rhine has been called to account for dropping percipients who no longer make better than average scores. But the dropping occurred only after the loss in telepathic or clairvoyant power persisted. We must revert to our sharpshooting example. If one member of a rifle team loses his skill, and it is skill that we are studying, why should he be retained? Rhine did not throw out scores when they did not happen to please him, as some of his critics have charged. All scores were lumped together to arrive at a total deviation, and the high ones reduced to a common average. Rhine has also been criticised on the ground that he told his subjects what their scores were after each series of 25 calls. This is exactly like telling a

team of riflemen how many hits and misses they made after each series of 25 shots. If the development of skill is one object—and it was one object in Rhine’s work—why shouldn’t the cardcallers know their scores? Finally there is also the objection that Rhine did not test his percipients with totally blank cards. He feared that trickery would destroy interest, which he had found to be essential. This observer thinks the experiment with blank cards might well be made. If clairvoyance is indeed a real gift, the percipient ought to “ see ” at least the image of a white card.

What impresses in Rhine’s work _is the non-statistical fact that when a list of calls of a target pack is compared with the order of cards in a shuffled pack the average of agreement is approximately one in five—just what would be expected. If the precipienta make more hits when compared with the target deck than with another pack they have obviously succeeded more or less in doing what they tried to do.; This is exactly the opposite of what is called a chance occurrence.

In addition, we have the remarkable fact that sleeping drugs (sodipm amytal, for example) reduce the high scores to what they should be on » chance basis, that caffeine raises scores, that some • percipients can make high and low scores at will. t . If the statisticians assure us that Rhine’s results are not merely chance coincidences, what have the psychologists to say? As yet nothing but a questioning of statistical methods. So we find ourselves spinning in circles. There is something to explain, and it is for the psychologist to explain it, instead of arguing about a mathematical procedure which his mathematical better! have pronounced good enough for lU purpose.