The Effects of Experimenters' Expectations

 Self-fulfilling prophecies

Frequently things turn out just as expected or prophesied, not because of a mysterious knowledge of the future, but because people's behaviour tends to make the prophecy come true. For example, a teacher who predicts that a student will fail may treat the student in ways that make failure more likely, thus fulfilling the original prophecy. The tendency for prophecies to be self fulfilling is well known in the realms of economics, politics and religion. It is also a matter of practical psychology. Various ways of using these powers are the bases of countless self-help books, showing how avoiding negative attitudes and adopting positive ones help to bring about remarkable successes in politics, business and love. Likewise confidence and optimism play an important part in the practice of medicine and healing - and in sports, fighting and many other activities.

However one chooses to interpret it, positive and negative expectations often influence what actually happens. Self fulfilling prophecies are commonplace. So how does this apply to science? Many scientists carry out experiments with strong expectations about the outcome, and with deep-rooted assumptions about what is and what is not possible. Can their expectations influence their results? The answer is yes.

First, expectations affect the kinds of questions that are asked in experiments. And these questions in turn shape what kinds of answers will be found. This is explicitly acknowledged in quantum physics, where the design of the experiment determines what kind of outcome is possible; for example, whether the answer will be in wave or particle form. But this principle is perfectly general. "The structure of the examination is like a stencil. It determines how much of the total truth will appear and what pattern it will suggest."

Second, experimenters' expectations affect what they observe, giving a tendency to see what they want to see and to ignore what they do not want to see. This tendency can lead to unconscious biases in observation and in the recording and analysis of data, to the dismissal of unfavourable results as errors, and to a very selective publication of results, as I discussed in the introduction to Part 3.

Third, and more mysteriously, experimenters' expectations may affect what actually happens. Just how mysterious this process might be is the question this chapter explores.

 Experimenter effects

 A pioneering piece of industrial research, carried out at the Hawthorne plant of the Western Electric Company in Chicago in 1927-9, has become familiar to generations of students of social psychology. It revealed what is now generally known as the 'Hawthome effect'. This study was designed to find out the effects on productivity of various changes in rest periods and refieshments. But, to the surprise of the investigators, output increased by about 30 per cent irrespective of the particular experimental treatments. The attention they were being given had a greater effect on the workers than the particular physical conditions they were working under.

The Hawthorne effect may play a part in many kinds of research, at least in psychology, medicine and animal behaviour. Investigators affect the subjects of their investigation merely by paying attention to them. Moreover, they may not only have a general influence, owing to their attention and interest, but also a specific influence on the way their subjects behave. In general, subjects tend to behave in accordance with experimenters' expectations.

The tendency for experiments to yield the expected results is known as the 'experimenter effect', or more precisely the experimenter expectancy effect'. Most researchers in the behavioural and medical sciences are well aware of this tendency, and try to guard against it by the use of 'blind' methodologies. In 'single-blind' experiments, the subjects do not know what treatment they are being subjected to. In 'doubleblind' experiments, the experimenters do not know either. The treatments are coded by a third party, and the experimenter does not know the code until after the data have been collected.

Important though experimenter effects are in research on human beings and animals, no one knows how widespread they are in other fields of science. The conventional assumption is that expenmenter effects are widely enough recognized already, and are confined to animal behaviour, psychology and medicine. They are largely ignored in other areas ofscience, as can easily be seen by visiting a scientific library and looking through the journals in different fields. In research in biology, chemistry, physics and engineering double-blind methods are rarely, if ever, employed. Scientists in these fields are generally innocent of the possibility that experimenters might unconsciously affect the systems they are studying.

Lurking in the background is the alarming thought that much of established science may reflect the influence of the experimenters' expectations, even through psychokinetic or other paranormal influences. These expectations may not only include those of individual investigators, but also the consensus among their peers. Scientific paradigms, models of reality shared by professionals, have a great influence on the general pattern of expectation and could innuence the outcome of countless experments.

It is sometimes suggested, in a joking way, that nuclear physicists do not so much discover new subatomic particles as invent them. To start with, the particles are predicted on theoretical grounds. If enough professionals believe they are likely to be found, costly accelerators and colliders are built to look for them. Then, sure enough, the expected particles are detected, as traces in bubble chambers or on photographic films. The more often they are detected, the easier they become to find again. A new consensus is established: they exist. The success of this investment of hundreds of millions of dollars then justifies yet further expense on even bigger atom smashers to find yet more predicted particles, and so on. The only limit seems to be set not by nature herself, but by the willingness of the US Congress to go on spending billions of dollars on this pursuit.

In the physical sciences, although there has been very little empirical research on experimenter effects, there have been many sophisticated discussions of the role of the observer in quantum theory. Such observers, discussed philosophically, sound like the detached minds of idealized objective scientists. But if the active influence of the experimenter's mind is taken senously, then many possibilities open up - even the possibility that the observer's mind may have psychokinetic powers. Perhaps 'mind over matter' phenomena take place in the microscopic realm of quantum physics. Perhaps the mind can influence the probabilities of happenings which are 'probabilistic', not rigidly determined in advance. This idea is the basis of much speculation among parapsychologists, and is one way of trying to explain the interaction of mental and physical processes in the brain.

In the realm of animal behaviour, as I discuss below, there is actual experimental evidence for the effects of experimenters' expectations on the behaviour of animals. But in most areas of biology, the possibility of such effects is usually ignored. An embryologist, for example, may well recognize the need to guard against biased observation and to use appropriate statistical procedures, but is unlikely to take seriously the idea that his expectations can, in some mystenous way, influence the development of embryonic tissues themselves.

In psychology and medicine, experimenter effects are generally explained in terms of influences transmitted by 'subtle cues'. But just how subtle these cues may be is another question. It is generally assumed that they depend only on recognized forms of sensory communication, in turn dependent only on the well-known principles of physics. The possibility that they include 'paranormal' influences such as telepathy and psychokinesis is not discussed in polite scientific society. I believe that it is better to face this possibility than to ignore it, and propose an investigation of experimenter effects that takes into account the possibility of 'mind over matter' effects. But first it is important to consider what has already been established.

How people behave as expected

 People generally behave as expected. If we expect people to be friendly they are more likely to be so than if we expect them to be hostile, and treat them accordingly The patients of Freudian analysts tend to have Freudian dreams, while patients ofJungian analysts haveJungian dreams. There are countless examples from all realms of human experience that illustrate this principle.

Compared with the richness of personal expenence and anecdotal accounts, experiments on the effects of expectation on people's behaviour seem contrived and trivial. Nevertheless, they are important in that they enable this effect to be investigated empirically and brought within the realm of scientific discourse. And hundreds of experiments have indeed shown that experimenters can affect the outcome of psychological investigations, biasing them in the direction oftheir expectations.

Here is one example. A group of fourteen psychology graduate students was given 'special training' in a 'new method of learning the Rorschach procedure', in which they would be asking people what patterns they saw in ink blots. Seven of them were led to believe that experienced psychologists obtained more human than animal images from their subjects. The other seven were given the same ink blots but told that they had been found by experienced psychologists to give rise to a higher proportion of animal images. Sure enough, the second group obtained significantly more animal images than the first.

Less trivial is the empirical demonstration that the effects of such expectations are not confined to short-term laboratory experiments. In schools, for instance, the way teachers treat pupils and hence the way the children learn is strongly influenced by expectations. The textbook example is called the 'Pygmalion experiment', carried out in an elementary school in San Francisco by the Harvard psychologist Robert Rosenthal and his colleagues. These prestigious scientists created expectations in the teachers that certain children in their classes were about to bloom intellectually and would show remarkable gains in the current school year. The psychologists created this belief by administering a test to all the children in the school, describing it as a new technique for predicting intellectual 'blooming', calling it the 'Harvard Test of Inflected Acquisition'. Within each class, the teacher was then given the names of the 20 per cent of the children who had scored highest. In fact, it was an ordinary non-verbal intelligence test, and the names of those most likely to 'bloom' were chosen at random.

At the end of the school year, when all the children were tested again with the same intelligence test, in the first grade, the 'promising' children scored an average of 15.4 IQ points more than the control children; in the second grade 9.5 points more. Not only did these 'promising' children tend to score better, but there was also a tendency for teachers to rate them as more appealing, adjusted, affectionate, curious and happy. This effect showed up much less from the third grade upwards, probably because the teachers had their own expectations about the children; the expectations created by Rosenthal and his colleagues had much less effect when they had to compete with established reputations. Many subsequent studies have confirmed and amplified these conclusions.

A criticism levelled against Rosenthal and his colleagues was that their own strong commitment to finding experimenter effects had biased their own results. Rosenthal replied that if this were so it would merely prove his point in another way:

 We could perform a study in which we randomly divided expectancy investigators into two groups: in the first, expectancy experiments would be conducted as usual, while in the second, special safeguards would be employed so that the expectancy of the principal investigator could not be communicated to the experimenters. Suppose that the average expectancy effect for the first group was 7, and for the second, zero. We would still view this as evidence for the phenomenon of expectancy effects!

 Although in the medical and behavioural sciences double-blind procedures are routinely employed to guard against experimenter effects, these methods are only partially effective. Some effects of expectancy still persist, and are most clearly seen in the placebo effect in medical research.

The placebo effect

 Placebos are treatments with no specific therapeutic value which nevertheless help to make many people better. Medical researchers have found that placebo effects are all-pervasive in medicine. If placebo effects are not controlled in therapeutic studies, the findings are generally considered unreliable. Placebo effects have been found in many conditions, including cough, mood changes, angina pectoris, headache, seasickness, anxiety, hypertension, status asthmaticus, depression, common cold, lymphosarcoma, gastric secretion and motility, dermatitis, rheumatoid arthritis, fever, warts, insomnia and pain symptoms from a variety of sources.

Much of the success of therapy through the ages can be attributed to the placebo effect, irrespective of the kind of therapy, or of the the ones supporting it. And there can be no doubt that it plays a large role in modern medicine as well. A survey of a wide range of drug trials has revealed that placebos are, on average, about a third to a half as effective as specific medication - a big effect for blank pills that cost almost nothing. But placebos are not just blank pills. They can also be forms of blank counseling or psychotherapy, or even blank surgery. For example, one surgical procedure for the treatment of angina pain involved the binding of the mammary arteries. When the effectiveness of this procedure was tested, the appropriate incision was made in control patients, but no artery was bound. 'Relief from the angina pain was the same among the real and sham surgery groups. In addition, both groups showed physiological changes, including reduction in the inverted T-wave of the EKG recording.'

So what are placebos? The history of the word itself is revealing. It is the first word of a chant in medieval funeral rites, 'placebo domino' - I shall please the Lord. The word was used to refer to professional mourners who were paid to 'sing placebos' at the bier of the deceased in place of the family, whose role it was originally. Over the course of several centuries the connotations of the term gradually became derisive; it was used to describe flatterers, sycophants and social parasites. It first appeared in a medical dictionary in 1785, in a pejorative sense, defined as 'a commonplace method or medicine'.

The professional placebo singers in the Middle Ages no doubt tended to lack any specific devotion to the deceased. Nevertheless their chanting was generally believed to be of value as part of an acknowledged ritual. Modern placebos are given in a therapeutic setting, and also depend for their power on current beliefs and expectations, both of the doctor and the patient. Any method of treatment in any culture, traditional or modern, occurs in a context in which the particular techniques are viewed by the patient as plausible and the therapists as potentially effective.

Doctors are often quick to ascribe the efficacy of traditional or 'unscientific' medical systems to the placebo phenomenon, and also to impute the use of placebos to other kinds of physician. But they tend to exempt their own kind of medicine. In one survey of attitudes to placebo effects, surgeons excluded surgery, internists excluded medication, psychotherapists excluded psychotherapy, and psychoanalysts excluded psychoanalysis. Moreover, in medical research, placebo effects are generally regarded as a nuisance. But perhaps the negative attitudes of physicians to placebos is just as well, since it is the other side of the coin of their faith in the special efficacy of their own techniques, which therefore tend to work better - because of the placebo effect!

The largest placebo effects occur in double-blind trials in which both patients and physicians believe a powerful new treatment is being used. If the treatment is believed by the doctors to be less effective, a smaller placebo effect is obtained. In single-blind trials, in which the doctors know which patients have been given the placebo but the patients do not, placebos are still less effective. In open conditions where the patients know they are receiving placebos, the effects are smallest of all. In other words, treatments work best if they are thought to have powerful benifical effects by both doctors and patients. Conversely, in trials where the active medications are labelled as placebos, the drugs give poorer clinical results.

Thus lowered expectations lead to a lowered placebo effect. This is the case with new 'wonder drugs' that arouse high hopes to start with, but fail to live up to expectations. This pattern was recognized by the nineteenth-century French physician Armand Trousseau, who advised his colleagues to 'treat as many patients as possible with the new drugs while they still have the power to heal'. There are many modern examples. For instance, at one time the drug chlorpromazine was hailed for its efficacy in treating schizophrenia, but then faith in its powers gradually waned. In successive trials it was found to be less and less effective. The effects of placebos declined in parallel. 'Perhaps as the investigators began to realize that the new "wonder drug" was not as powerful as they had hoped, their expectations, and possibly their interest in the patients declined.' Here is another particularly striking example, from the 1950s:

 A man with advanced cancer was no longer responding to radiation treatment. He was given a single injection of an experimental drug, Krebiozen, considered by some at the time to be a 'miracle cure' (it has since been discredited). The results were shocking to the patient's physician, who stated that his tumors 'melted like snowballs on a hot stove'. Later the man read studies suggesting the drug was ineffective, and his cancer began to spread once more. At this point his doctor, acting on a hunch, administered a placebo intravenously. The man was told the plain water was a 'new, improved' form of Krebiozen. Again, his cancer shrank away dramatically. Then he read in the newspapers the American Medical Association's official pronouncement: Krebiozen was a worthless medication. The man's faith vanished, and he was dead within days.

 The same principles apply to medical research itself. Believers and non-believers in new forms of treatment tend to obtain very different results: 'Quantitatively, the pattern is consistent. The initial 70 to 90 percent effectiveness in the enthusiasts' reports [decreased] to 30 to 40 percent "baseline" placebo effectiveness in the skeptics' reports.'

A remarkable feature of placebos is that patients not only benefit from them, but also exhibit toxic responses or side effects. In one survey of sixty-seven double-blind drug trials involving 3,549 patients, 29 per cent of the patients showed various side-effects while they were being treated with the placebo, including anorexia, nausea, headache, dizziness, tremor and skin rash. The side-effects were sometimes so severe that they required additional medical intervention. Moreover, they showed a relationship to the doctors' or patients' expectations about the active drug being used in the trial. For example, in a large-scale double-blind trial of oral contraceptives, 30 per cent of the women who were administered the placebo reported decreased sex drive, 17 per cent increased headache, 14 per cent increased menstrual pain and 8 per cent increased nervousness and initability.

Just as the power of blessings is mirrored by the power of curses, so the beneficial effects of placebos are mirrored by the negative effects of procedures expected to bring about harm, technically known as 'negative placebos' or 'nocebos'. Spectacular examples in Africa, Latin America and elsewhere are known to anthropologists as 'voodoo deaths', brought about by belief in the power of bewitchment. Less spectacular nocebo effects have also been demonstrated in laboratory experiments, as in a study in which subjects were told that a mild electric current was being passed through their head by means of applied electrodes, and warned that this might give rise to a headache. Although there was in fact no current, two-thirds of the subjects developed headaches. Both placebos and nocebos depend on prevailing cultural beliefs, including the belief in scientific medicine. 'Simply put, belief sickens; belief kills; belief heals.'

The influence of expectancy on animals

 Animals respond to different people differently, as every pet owner and animal trainer knows. They recognize people they are used to, and tend to be on their guard with strangers. They seem to sense whether people are friendly, gauge their fear or confidence, and respond to their expectations. From a common sense point of view, based on everyday experience, it is hardly surprising that scientists who do experiments on animals have a personal influence on the animals. The experimenters' attitudes and expectations affect the animals they work with.

The classical experiments on the effects of experimenters' expectations on animals were carried out in the 1960s by Robert Rosenthal and his colleagues. They used students as experimenters and rats as subjects. The rats came from a standard laboratory strain, but were divided at random into two groups, labelled 'Maze-Bright' and 'Maze-Dull'. The students were told that these animals were the products of generations of selective breeding at Berkeley for good and poor performance in standard mazes. The students naturally expected the 'bright' rats to learn quicker than the 'dull' ones. Sure enough, this is what they found. Overall the 'bright' rats made 51 per cent more correct responses and learned 29 per cent faster than the 'dull' rats.

These findings have been confirmed in other laboratories and with other kinds of learning. Comparable expenmenter effects have even been observed with flatworms, lowly creatures that live in mud at the bottom of ponds and in similar aquatic environments. In one such study, a sample of essentially identical Planaria worms was divided into two groups, one of which was described as a strain showing few head turns and body contractions ('low-response-producing worms'), and the other as a frequent turner and contracter ('high-response-producing worms'). With these expectations in mind, the experimenters found on average five times more head turns and twenty times more contractions in the 'high response-producing' worms.

These expectancy effects, like those in Rosenthal's rat experiment, were shown by undergraduate students, who may be especially prone to see, or even to pretend to see, what they are told to expect. Seasoned observers might generally show smaller expectancy effects. This was the case, for example, when more-experienced researchers were working with Planaria. The number of contractions in 'high-response-producing' Planaria was found to be two to seven times greater than in 'low response-producing' worms, compared with the average of twenty times greater found by undergraduates. Nevertheless, a two- to sevenfold increase is still a large effect, and obviously introduces a serious bias into the results.

On the other hand, experienced observers may be strongly committed to particular views of reality, directly or indirectly resulting in greater expectancy effects than those found among novices with less personal commitment to particular theories. They may create a climate of expectation among their colleagues and technicians, and this in turn may influence the way their animals behave.

Although expectancy effects were first systematically investigated in the 1960s and have now been demonstrated in hundreds of special studies, the general principle is by no means new. For example, Bertrand Russell, writing with his customary wit and clarity, put it as follows in 1927:

 The manner in which animals learn has been much studied in recent years, with a great deal of patient observation and experiment.... One may broadly say that all the animals that have been carefully observed have behaved so as to confirm the philosophy in which the observer believed before his observations began. Nay, more, they have all displayed the national characteristics of the observer. Animals studied by Americans rush about frantically, and with an incredible display of bustle and pep, and at last achieve the desired result by chance. Animals observed by Germans sit still and think, and at last evolve the solution out of their inner consciousness.

Experimenter effects in parapsychology

 Experimenter effects are well known to parapsychologists, for several reasons. First, it has long been known to experienced researchers that subjects tend to show more psychic powers when they are feeling relaxed, and in a positive and enthusiastic atmosphere. If they are anxious, uncomfortable or treated in a formal and detached way by the scientific investigators, they do not perform so well. In fact they may show no significant psychic powers at all, no 'psi effects', in the jargon of parapsychology.

Second, it is a matter of common observation among researchers in this field that subjects who show considerable psychic abilities often tend to lose them when strangers come into the room as observers.The pioneering parapsychologst J.R. Rhine actually quantified this effect in a series of trials with a gifted subject, Hubert Pearce, having noticed that when someone called in to see Pearce at work his scores at once dropped down. 'We began to take down evidence, sometimes inviting a visitor for that purpose, sometimes availing ourselves of a casual caller. We recorded the time of entrance and exit on 7 visitors, one beiIlg present twice. They all produced a drop in Pearce's scoring.'

The off-putting effect of strangers is particularly strong when the strangers are sceptical, especially if they are hostile to the experiment itself or to the people involved. However, if strangers are friendly, and especially if they help in some way in the experiment, rather than behaving as detached observers, subjects get used to them and psi scores rise again.

Sceptics usually take the failure of parapsychological tests in the presence of sceptics to learn that psychic powers cannot be detected under rigorous scientific conditions, and therefore don't really exist. But the negative effects of sceptics may well be due to their off-putting presence and negative expectations, mediated by subtle and not-so-subtle cues.

Third, it is well known among parapsychologists that some experimenters consistently obtain positive results in their research, while others do not. This effect was systematically investigated in the 1950s by two British researchers. One, C.W. Fisk, a retired inventor, consistently obtained significant results in his experiments. The other, D.J. West, later to become Professor of Criminology at Cambridge, was usually unsuccessful in detecting psychic phenomena. In these experiments, each investigator prepared half of the test items, and scored them at the end. The subjects did not know that two experimenters were involved, nor did they meet them; they received the test items through the post and also returned them by mail. The results from Fisk's half of the experiment showed highly significant effects for clairvoyance and psychokinesis. West's data showed no deviation from chance. They concluded that West was 'a jinx'. 

Fourth, in research on psychokinesis it has repeatedly been found that experimenters who find significant effects are themselves good subjects. For example, Helmut Schmidt, the inventor of the Schmidt machine, a random number generator whose output can apparently be affected by willing certain patterns to emerge, has found that he is often his own best subject. One investigator, Charles Honorton, has even shown that psychokinetic effects on random number generators by the subjects in his experiments are more due to himself than to his subjects. The subjects showed psychokinetic powers when he was present; and he himself showed them when he was acting as the experimental subject. But the psi effect was lost when he was not present and the subjects were tested by another experimenter. Honorton and his colleague Barksdale concluded that such effects showed that 'traditional boundaries between subjects and experimenters cannot be easily maintained'. They interpreted their results as a psi-mediated experimenter effect'.

The implications of such experimenter effects are staggering. If parapsychologists can bring about psi-mediated experimenter effects, whether intentionally or not, through their influence over their subjects, even at a distance (as in the Fisk-West experiments), then the conventional separation between experimenters and the subjects of their investigation breaks down. Moreover, if people can influence physical events such as radioactive decay, then the conventional separation between mind and matter breaks down too. But then why should psi mediated experimenter effects be confined to parapsychology? Might they not occur in many other fields of science?

How paranormal is normal science?

There is a good reason for the conventional taboo against parapsychology, making it a kind of outcast from established science. The existence of psychic phenomena would seriously endanger the illusion of objectivity. It would raise the possibility that many empirical results in many fields of science reflect the expectations of the experimenters through subtle unconscious influences. Ironically, the orthodox ideal of passive observation may well provide excellent conditions for paranormal effects:

 An experimenter preparing his apparatus, getting his animals ready, and then leaving them with some feeling of assurance that the experiment will run and the animals will appropriately 'do their thing' cannot but remind us of certain aspects of magic, ritual, or perhaps petitionary prayer. Something is done with the confidence that it will produce a desired result, and the participant, once he has done this, psychologically puts a distance between himself and the outcome. He is not trying to make things happen, but just trusts that they will.... Such circumstances may provide all optimum opportunity for psychokinetic intervention.

 This possibility has indeed been raised in a paper in Nature entitled 'Scientists confronting the paranormal', by the physicist David Bohm and others. They noted that the relaxed conditions necessary for the appearance of psychokinetic phenomena are also those most fruitful for scientific research in general. Conversely, tension, fear and hostility tend not only to inhibit psi effects, but also to influence experiments in the so-called hard sciences too. 'If any of those who participate in a physical experiment are tense and hostile, and do not really want the experiment to work, the chances of success are greatly diminished.'

The defenders of orthodoxy generally reject or ignore the possibility of paranormal influences under any circumstances. The task of keeping science psi-free is undertaken by organized groups of Skeptics. These scientific vigilantes continually challenge any evidence for psi effects, rejecting it on one or more of the following grounds:

 1 Incompetent experimentation.

2 Selective observation, recording and reporting of data.

3 Unconscious or conscious deception.

4 Experimenter effects mediated by subtle cues.

 Skeptics are right to point out these possible sources of error in parapsychological research. But the same sources of bias are present in orthodox research as well. The very fact that parapsychological research is subject to such cntical scrutiny makes researchers in this field unusually conscious of the effects of expectation. Ironically, it is in conventional, uncontroversial fields of research that the influences of experimenters' expectations are most likely to pass undetected.

The evidence for experimenter effects in medicine and the behavioural sciences is undeniable. And that is why 'subtle cues' take on such an important explanatory role. Almost everyone agrees that subtle cues such as gestures, eye movements, body posture and odours can influence people and animals. Sceptics are very keen on emphasizing the importance of such cues, and rightly so. A favourite example showing the importance of subtle communication is the story of Clever Hans, a famous horse in Berlin at the turn of the century. This horse could apparently perform arithmetic in the presence of its owner by tapping a hoof on the ground to count out an answer. Fraud seemed unlikely, since the owner would allow other people (free of charge) to question the animal themselves. The phenomenon was scientifically investigated in 1904 by the psychologist Oskar Pfungst, who concluded that the horse was receiving clues from gestures made, probably unwittingly, by the owner and other questioners. Pfungst found that he could get the horse to give the correct answer simply by concentrating his attetion on the number, though he was not aware of making any movement that would give the number away.

No one denies that subtle cues from experimenters, passing through normal sensory channels, can affect people and animals. Sceptics claim that such influences may explain many examples of seemingly telepathic communication. But granted all this, the possibility remains that both subtle sensory cues and 'paranormal' influences play a part.

The story of Pfungst's investigation of Clever Hans has been told again and again to generations of psychology students. What is less well known is that after Pfungst's investigation, described in his book on Clever Hans published in 1911, further studies on horses with similar mathematical powers showed that more was involved than subtle sensory cues. For example, when Maunce Maeterlinck investigated the famous calculating horses of Elberfeld, he concluded that they were somehow reading his mind, rather than responding to subtle sensory cues. After a series of increasingly stringent tests, he finally thought of one which 'by virtue of its very simplicity, could not be exposed to any elaborate and farfetched suspicions'. He took three cards with numbers on them, shuffled them without looking at them, and placed them face down on a board where the horse could see only their blank backs. 'There was therefore, at that moment, not a human soul on earth who knew the figures'. Yet, without hesitation, the horse rapped out the number the three cards formed. This experiment succeeded with the other calculating horses too 'as often as I cared to try it'. These results go even beyond the possibility of telepathy, since Maeterlinck himself did not know the answers when the horses were tapping them out. They imply either that the horses were capable of clairvoyance, directly knowing what was on the cards, or precognition, knowing the number that would be in Maeterlink's mind when he later turned the cards over.

 For more than eighty years, the story of Clever Hans and Pfungst has been told and re-told as a triumph of scepticism. It has taken on a mythic significance, enabling seemingly paranormal effects to be explained in terms of subtle cues. But what if some of the subtle cues are themselves paranormal? There is a taboo against even discussing this possibility, let alone investigating it. Nevertheless, the possible importance of parapsychological influences was suggested to Rosenthal by one of his colleagues at Harvard right at the outset of his research on experimenter effects:

 Had I the wit or courage to do so, I could easily have conducted a study in which experimenters with varying expectations for their subjects' responses were prevented from having sensory contact with those subjects. My prediction, then and now, was (and would be) that under these conditions no expectancy effects could occur. But I never did the study.

 Maybe if someone actually did this study, Rosenthal's prediction would turn out to be wrong. Maybe some of the effects of experimenters' expectations are indeed paranormal. Such subtle influences would not be opposed to subtle cues; they would usually work along with them, and operate just as unconsciously.

Although experimenter effects are well recognized in the medical and behavioural sciences, the fact that they are explained - or explained away - in terms of 'subtle cues' prevents them being taken very seriously in other fields of investigation such as biochemistry. Whereas a person or a rat might pick up a scientist's expectations and respond accordingly, an enzyme in a test-tube would not be expected to respond to subtle body language, unconscious facial gestures, etc. Of course, there is a general recognition of the possibility of biased observation, but this is not a result of any actual influence on the experimental system itself. The scientist may 'see' a difference that fits his or her expectancy, but the difference is supposed to be only in the eye of the observer, not in the material studied.

Nevertheless, all this is merely an assumption. There has been practically no research on the influence of experimenters' expectations in fields of science such as agriculture, genetics, molecular biology, chemistry and physics. Since the material studied is assumed to be immune from such influences, precautions against them are assumed to be unnecessary. Except in the behavioural sciences and in clinical research, double-blind procedures are rarely employed.