Robert Thouless

Dr. Robert H. Thouless

Educated at Corpus Christi College, Cambridge, where he was awarded his PhD in 1922. He then went on to become a Lecturer in Psychology at Manchester, in Glasgow, and again in Cambridge. Here he became the Reader in Educational Psychology. He was President of the British Association's Psychology Section in 1937, and published a number of books connected with this subject. President of the SPR from 1942 till 1944.

Can the Experimenter be Trusted?

 - Robert H. Thouless -

          THE PRECAUTIONS described in the last chapter are those necessary for the experimenter to satisfy himself that his experiments, if successful, are indicating the operation of paranormal capacities on the part of his experimental subject. They are not in principle different from the kind of precautions that would be taken in any other branch of experimental science, as for example, in an experiment on visual acuity in a psychological laboratory. Let us call such precautions primary precautions. If full primary precautions are taken by an experimenter and if also his experimental subject is succeeding too often for his successes to be reasonably attributed to chance, then the experimenter can be satisfied that the experimental subject has whatever paranormal capacity (ESP, precognition, or PK) the design of the experiment was adequate to reveal.

The experimenter can be so satisfied but not necessarily anyone else. All that is normally required in an experiment is that its conditions should be such as to satisfy the experimenter, and for most experiments in parapsychology, the primary precautions are enough. They are not, however, enough if the experimenter has set himself the task of producing evidence that will be completely convincing to other people. These others may have varying degrees of unwillingness to believe in any results which seem to favour a paranormal hypothesis. Some of them may argue that ESP is impossible and that dishonesty on the part of the experimenter (or of one or more of his helpers) may be admitted to be very improbable but is plainly not impossible. So being faced with a choice between an impossible and an improbable explanation, we must choose the improbable one.[1] In any experiment with full primary precautions the experimenter can know that he has not cheated; he may try to devise a system of secondary precautions so that others can know it too.[2]

[1] George R. Price, 'Science and the Supernatural', Science, 26 August 1955 (abstracted in the Journal of Parapsychology, XIX (Dec. 1955), pp. 238-41).
[2] This is, of course, a special type of experiment; most parapsychological experiments have some other object than that of convincing other people that the experimenter did not fake his results. It is a common misunderstanding of the situation to suppose that all experiments should conform to the design dictated by this special requirement and that those which do not are therefore imperfect. If we require coercive proof of the reality of ESP by means fully safeguarded against the possibility of fraud by the experimenter (or by any other person helping to conduct the experiment), this requirement would be met by only one experiment that had been so safeguarded and had been successful. If there is already more than one experiment that has been so safeguarded, it would be a waste of time to do any more of the same kind. On the other hand, if all existing experiments fall short of perfect safeguarding against experimenters' fraud because they all can be shown to have imperfections (ways in which, whether he did or not, the experimenter could have cheated), and if improved designs of experiment would get rid of those imperfections, then it would seem that the sensible thing would be to carry out those new experiments with improved design.

It may be that neither of these views about the experimental situation is the true one. My own opinion is different from both; it is that a completely conclusive experiment of this type is, in principle, impossible. I do not think this impossibility is peculiar to parapsychology: it would apply equally to an experiment in any other field in which the untrustworthiness of the experimenter was seriously considered as an explanation of results. It does not trouble us in other fields because generally their experiments are more easily repeatable than those of parapsychology. It would be an exaggeration to say that parapsychological experiments are not repeatable; they are constantly being repeated. There is no certainty that a new investigator repeating an investigation will get the same result but a considerable probability that he will. If the strength of experimental evidence for the reality of paranormal events is to be increased, there is better hope of this end being achieved by the multiplication of successful experiments with adequate primary precautions than by the pursuit of the single experiment with a water-tight system of precautions against experimenters' fraud.

Even if this be accepted, it remains an interesting question whether the experiments that have already been done with secondary precautions against fraud by the experimenter or his colleagues are as good as they can be made or whether new experiments of this type should be designed with new and better precautions. Whatever we may think of their finality, they remain an important part of the case for the reality of ESP. It remains a strong argument for the reality of paranormal phenomena that various experiments force as to choose between acceptance of their reality and the very unlikely alternative of a conspiracy to deceive entered into by a number of persons who seem honest and reliable. This argument remains strong even if it is not absolutely coercive against those who accept as axiomatic the proposition that paranormal phenomena are impossible.

In experiments with this aim, there have been various ways adopted for controlling the experimenter. The record that the experimenter makes of the target series of cards may be handed to a third person or group of persons before the experimenter knows what the percipient has guessed. This same person or group of persons also may receive the guessed series before the percipient knows what the target series was, and the comparison between the two series may be made by this third person or group of persons. Then if adequate precautions have been taken to see that the guesses are those which were made by the percipient and that the target series represents the actual order in which the cards were presented by the agent, and if also an independent witness has shuffled the target set of cards and has watched to see that the cards so shuffled are the ones used in the experiment, then one has ensured that successful results could not have been fraudulently produced by the experimenter himself. It would still be possible that they could be fraudulently produced by the experimenter in collusion with all or with some members of the witnessing group. This possibility can be made negligibly small by increasing the reliability of the witnesses and the number of reliable witnesses who have had complete control for some part of the experimental series. But the possibility does not become zero; it is still reasonable for the critic who rejects the possibility of ESP to say that if no other normal explanation remains open, there must have been collusive fraud to produce the results.

Instead of having records of targets and guesses delivered to some person or group outside the experiment, the same object (of making it impossible for the experimenter to cheat) may be attempted by having the recording of targets and guesses and the process of comparing them witnessed throughout by outside witnesses. These may be different persons on different occasions and if success is repeated on different occasions, it becomes difficult to suppose that all these witnesses have joined a conspiracy to mislead the public. Again the critic may say: very difficult but not impossible, and what the experiments seem to prove are impossibilities.

In the early days of experimenting in ESP, it was not felt necessary to take elaborate precautions against the experimenters cheating. Many experiments with this aim have, however, now been carried out; I will mention three which are typical and are amongst those most discussed. The earliest was the Pearce-Pratt series of experiments carried out at Duke University in 1933-4.[3] Here Dr Pratt was the agent, while a divinity student called H. E. Pearce was the percipient. There were four sets of experiments totalling 74 runs through the pack. Since there were twenty-five cards in each pack, the total number of the subject's guesses was 1,850, of which one would expect 1/5th (that is 370) to be right by chance. The number of right guesses actually scored was 558 which is about 50 per cent more than the 370 to be expected by chance. The percipient was sent by himself to the library and the agent was in another building, 100 yards away in some experiments and 250 yards in others. The precaution taken against cheating by percipient or agent or by both in collusion was that the records of targets and of guesses were handed in a sealed packet to Dr Rhine immediately after the experiment before they were compared. This precaution does not, of course, eliminate the theoretical possibility that Pratt, Pearce, and Rhine might be in a joint conspiracy to deceive as to the results. It has also been suggested by Hansell[4] that the percipient might himself have produced the results dishonestly if he had left the library after he was seen to enter it and had managed to conceal himself somewhere where he could see the laying out of the target packs. It is obviously highly improbable that he would have tried to do this, and also that he could have succeeded in doing it without detection. Highly improbable but not impossible, so this series of experiments will not be sufficiently convincing to a critic who holds that the ESP explanation is impossible.

[3] J. B. Rhine and J. G. Pratt, 'A Review of the Pearce-Pratt Distance Series of ESP Tests', Journal of Parapsychology, XVIII (1954), pp. 156-77.
[4] C. E. M. Hansell, 'A Critical Analyses of the Pearce-Pratt Experiment', Journal of Parapsychology, XXV (1961), pp. 87-91.

Later at Duke University came the Pratt-Woodruff experiment.[5] Here also there were safeguards against the experimenter cheating, and this series also has been subjected to criticism.[6]

[5] J. G. Pratt and J. L. Woodruff, 'Size of Stimulus Symbols in Extra-Sensory Perception', Journal of Parapsychology, III (1939), pp. 121-58.
[6] C. E. Hansell, 'A Critical Analysis of the Pratt-Woodruff Experiment', Journal of Parapsychology, XXV (1961), pp. 99-113.

The most elaborately safeguarded experiment to date has been that of Dr Soal and Mrs Goldney carried out with the gifted percipient Shackleton in London in 1941.[7] These experimenters used pre-arranged lists of the digits 1-5 prepared from a table of random numbers. There were five target cards with animal pictures on them, one of which was exposed to the agent, in accordance with a code for translating the number on the list to one of the five animal representations. It was these animals (lion, elephant, etc) that Shackleton guessed. Various outside witnesses were brought in for each experiment to see that the experimenters were neither deliberately falsifying their results nor doing so accidentally by systematic recording errors. An earlier pilot experiment had shown that Shackleton seemed to be guessing, not the card turned up, but the card ahead of that, and the experiment was designed to discover whether this guessing one ahead would continue. Over a long series, he showed a very high rate of success, varying about eight right guesses per pack instead of the chance expectation of five. The rate of success was such that it could not reasonably be attributed to chance, so the critic who believes ESP to be impossible must consider the alternative hypothesis of deliberate fraud.

[7] K. M. Goldney and S. G. Soal, 'Experiments in Precognitive Telepathy', Proceedings of the SPR, XLVII (1943), pp.21-150.

In 1955, Dr G. R. Price of the University of Minnesota suggested that the Soal-Goldney results could have been produced by fraud if a sufficient number of the witnesses had conspired with the experimenter to produce the deception [see ref. [1]]. Later it was suggested by Hansell that the successful agents might have memorized the card order and produced successful results by changing the coding cards so as to produce maximum scores.[8]

[8] C. E. M. Hansell, 'A Critical Review of the Experiments with Mr Basil Shackleton and Mrs Gloria Stewart as Sensitives', Proceedings of the SPR, LII (1960), pp. 1-42.

These are perfectly proper criticisms; they are not suggestions that anyone has cheated but assertions that the conditions of the experiment were such as not to preclude the possibility of some individual or group of individuals having produced the results by fraud. The particular suggestions as to how fraud might have produced the results may be erroneous, for example, it has been objected that Hansell's suggestion could not account for as high a level of success as was obtained in the Soal-Goldney experiment. If this is admitted, the essentials of the criticism remain. 'The impossible does not happen. ESP is impossible. Therefore, ESP does not happen, and if any experimental results are such that they can only be explained by ESP or by fraud, the explanation by fraud must be accepted, however improbable it may appear.' The critic is naturally happier if he can suggest how the fraud took place, but it is not essential. If the above argument is accepted, fraud must have taken place, either in the way suggested or in some other way.

Obviously many people do not accept the argument. They will not deny the first premise, that the impossible does not happen. This indeed is a tautology which it would be nonsensical to deny. The second premise that ESP is impossible is not self-evident and may reasonably be doubted. We may indeed consider that our intuitive opinions as to what is and what is not possible are so unreliable that the only way to make sure whether any kind of event is possible is to find out empirically whether or not it happens. If it happens, it must be possible, whatever our intuitions may be in the matter; if it cannot be shown to happen then it may be impossible.

But for those who do accept as axiomatic the impossibility of ESP, the conviction that ESP does not happen cannot be overthrown by multiplication or refinement of the system of secondary precautions. Somewhat inconsistently Price suggested that a new ESP experiment should be done with the cards in welded steel containers and a jury of twelve strongly sceptical scientists. But even if such an experiment were successful, the critic who accepted the impossibility of ESP would still have to say that the twelve scientists must have entered into a conspiracy to defraud by some method unknown. The improbability would be great but it would still not be impossible. Even then we should have to prefer an improbability to an impossibility.

This being the case, we must consider whether there is not another road to certain conviction. Undoubtedly there is, and a good example of a sceptical psychologist taking that road, is to be found in the experiment of Professor Lucien Warner reported in 1937.[9] The one person who knows whether the experimenter has cheated is the experimenter himself. So if he carries out a successful experiment with full primary precautions and knows that he has not himself cheated, the evidence will be conclusive to him. A successful experiment means, of course, one too successful to be reasonably attributed to chance, and full primary precautions means full precautions against the various sources of error (sensory leakage, recording errors, etc.) and against the possibility of the subject or any other participant in the experiment producing the results fraudulently. Professor Warner used a percipient who had already shown his ability to succeed in ESP tests. He had the percipient in a locked room on a different floor from the experimenter with communication only by signal from percipient to experimenter and not the other way. In 250 guesses he got 93 hits, an excess of 43 over the 50 right guesses that would be expected if there were only chance correspondence between the target cards and the percipient's guesses. The odds against such an excess occurring by chance are of the order of a thousand million to one, so the result is convincing evidence of the percipient's guesses being determined by the target cards, and therefore of the reality of ESP; convincing evidence for Professor Warner himself and for anyone else who believes in Professor Warner's good faith. Not, of course, for the critic who believes that ESP is more unlikely than any alternative explanation; if he is to be convinced, it must be by doing a similar experiment himself.

[9] Lucien Warner, 'A Test Case', Journal of Parapsychology, I (1937), pp. 234-8.

It might be a good thing if parapsychological research laboratories sometimes refrained from the intensive experimenting with a successful percipient which seems to inhibit their ESP capacity. Instead some might remain as a pool of potential experimental subjects who could be made available for short series of experiments by sceptical critics of ESP who wished to expose themselves to the risk of being convinced by the only means conclusive against the alternative explanation of experimenter's fraud.

It may be that convincing those who deny the possibility of ESP is less important than is commonly supposed. If, however, the attempt is made, this seems a more promising way than the elaboration of secondary precautions. These are open to the objection of leading to unnecessary complications of design in experiment, and complications lead to new loop-holes. It is significant that Hansell's criticisms of the Soal-Goldney experiment suggest as the possible means of fraud a feature (the five cards used for coding) which was introduced to make it more difficult for the experimenters to cheat. New complications might be introduced to eliminate these possibilities but a new critic might come along to show how these have introduced new possibilities of error. I think it is better to stick to simple experimental designs, and to eliminate the theoretical possibility of experimenters cheating, not by devising elaborate precautions against this possibility, but by providing favourable opportunities for critics to see whether they can, as experimenters, repeat the reported successes.


"Experimental Psychical Research" by Robert Thouless (Middlesex, Baltimore, Victoria: Penguin Books, 1963).


More articles by Robert Thouless

The Experimental Study of Survival
The Problem of the Reality of ESP
The Critics of ESP
The Pattern of ESP
The Beginnings of Psychical Research
Anecdote and Experiment

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


Some parts of this page 2012