From Abracadabra to Zombies | View All
"...psi is a statistically significant departure of results from those expected by chance under circumstances that mimic exchanges of information between living organisms and their environment, provided that (a) proper statistical models and methods are used to evaluate the significance, and (b) reasonable precautions have been taken to eliminate sensory cues and other experimental artifacts." --John Palmer (1983)
"The term psi denotes anomalous processes of information or energy transfer, processes such as telepathy or other forms of extrasensory perception that are currently unexplained in terms of known physical or biological mechanisms. The term is purely descriptive: It neither implies that such anomalous phenomena are paranormal nor connotes anything about their underlying mechanisms." --Daryl J. Bem and Charles Honorton*
"Parapsychologists have no positive theory or model of the phenomena they claim to be studying. What they call “psi” is defined and identified negatively. They claim having demonstrated the existence of “psi” whenever they obtain a statistically significant result which cannot be readily explained by mundane causes. This strategy has many undesirable problems from a scientific perspective. For one thing, it is impossible to discover every possible normal cause in a particular experiment. For another, this allows them to claim any departure from chance as evidence for psi." --Ray Hyman*
The psi assumption has been made by many parapsychologists since the scientific investigation of psychic phenomena began in earnest with the establishment of the Society for Psychical Research in 1882. In the first scientific test of psychic power done for the society, Sir William Fletcher Barrett, a professor of physics at the Royal College of Science in Dublin, declared that he had evidence that the subjects of his study gave evidence of paranormal abilities because they could perform at guessing games significantly better than chance would predict. He did a number of experiments with the Creery sisters and their servant girl and came away declaring that the odds of their being able to guess correctly in one experiment “were over a million to one.” The odds of their guessing correctly five cards in row were “over 142 million to one” and guessing correctly eight consecutive names in a row were “incalculably greater” (Christopher 1970: 10). Of course, we now know the girls were able to do so well in the games because they were cheating. However, we should not make the mistake of thinking that there are only two possibilities when amazing feats at guessing games are achieved. Even if such subjects are not cheating, it does not follow that they are using psychic abilities.
Many of J. B. Rhine's experiments involved using a specially designed deck of cards developed by one of his assistants named Zener. A deck consists of 25 cards with 5 cards each of a star, three vertical wavy lines, a plus sign, a circle, and a square. If a subject correctly named 5 out of the shuffled deck of 25 ESP cards, it was considered pure chance. Certain subjects consistently named 6 out of 25 cards correctly. Rhine and his associates concluded that this departure from chance expectancy demonstrated the existence of ESP. He even concluded that subjects who consistently named 4 out of 25 cards also showed psychic ability. He called it psi missing and attributed it to the subject’s negative attitude toward both him and the paranormal. It is possible some of Rhine's subjects were cheating. We know that some of the decks of cards he used were transparent, allowing receivers to see what card the sender was looking at. It is also possible that the distribution of subjects who scored above, below, and at chance levels is exactly what would be predicted by the laws of probability.
Another example of the psi assumption can be found in the work of S. G. Soal (1889-1975), a mathematician at Queen Mary College, London University, who aimed to improve on Rhine’s rather sloppy methods by systematically excluding what sensory leakage (non-telepathic communication) and deception of any kind in his ESP experiments. By 1939 he had tested over 160 subjects for ESP in more than 128,000 guessing trials. He found no evidence of telepathy. Actually, what he found was nothing of statistical interest. That is, he found nothing of statistical interest until he went data mining for displacement. Soal found statistically significant numbers with two of his 160 subjects when he correlated guesses with cards preceding or following the target cards. He and others took this as evidence of clairvoyance. We now know that Soal didn’t go data mining. He went data changing (Hansel 1989: 111-116).
In any case, in addition to the problems of cheating by subjects and fraud by experimenters, there are two other kinds of problems with the psi assumption, one logical and one methodological. From a logical point of view, parapsychologists are either begging the question (assuming what they should be proving) or they are committing the fallacy of affirming the consequent. (If it’s psi, then the data deviate from chance. The data deviate from chance. So, it’s psi. Or, if a person is psychic, then that person will do better than chance in guessing experiments. That person did better than chance in a guessing experiment. Therefore, that person is psychic.)
The assumption is also questionable on methodological grounds. There is no reason to believe that the laws of probability, which are purely formal and ideal, should apply directly to any finite set of events. It may be true that the odds of a coin coming up heads or tails is 1 in 2, but that gives us no information as to what will happen in the real world for any given number of tosses. Ideally, in a large number of tosses, heads should come up 50% of the time. In the real world, there is no way to know exactly how many times heads will come up in, say, ten million tosses. We can be pretty sure the number will be very close to five million (assuming a fair coin and a fair toss), but we cannot know a priori exactly how many times heads will come up.
Studies comparing random strings with random strings, to simulate guessing numbers or cards, have found significant departures from what would be expected theoretically by chance (Alcock 1981: 159). For example, Harvie “selected 50,000 digits from various sources of random numbers and used them to represent “target cards” in an ESP experiment. Instead of having subjects make guesses, a series of 50,000 random numbers were produced by a computer.” He found a hit rate that was significantly less than what would be predicted by chance (Alcock 1981: 158-159).
In the 1930s, Walter Pitkin of Columbia University printed up 200,000 cards, half red and half blue, with 40,000 of each of the five ESP card symbols. The cards were mechanically shuffled and read by a machine. The result was two lists of 100,000 randomly selected symbols. One list would represent chance distribution of the symbols and the other would represent chance guessing of the symbols. However, the actual matches and what would be predicted by accepted odds didn’t match up. The total number was 2% under mathematical expectancy. Runs of 5 matching pairs were 25% under and runs of 7 were 59% greater than mathematical expectancy (Christopher 1970: 27-28). The point is not whether these runs are typical in a real world of real randomness or whether they represent some peculiarity of the shuffling machine or some other quirk. The point is that it is not justified to assume that statistical probability based on true randomness and a very large number of instances applies without further consideration to any finite operation in the real world such as guessing symbols in decks of 25 cards shuffled who knows how or how often, or rolling dice, or trying to affect a random number generator with one's mind. As Alcock put it: “If such significant variation can be produced by comparing random strings with random strings, then the assumption that any significant variation from chance is due to psi seems untenable (Alcock 1981: 158-159).”
The defender of psi might well ask: Why do some people perform better than chance in some experiments? We know some do better because they cheat. Some cheaters have admitted their cheating. Of course, it is possible that the Creery sisters and others who have admitted to cheating were lying about it but that seems farfetched. Of some others we might say we don't know for sure, but we have good reason to believe they cheated. For example, Hubert Pearce, Jr., one of J. B. Rhine's psychic stars, performed exceptionally well in card guessing experiments. In nearly 700 runs of the ESP cards, he averaged approximately 32% success vs. a chance expectation of 20%. But when Rhine used a magician to observe Pearce, he performed at chance levels. “There are at least a dozen ways a subject who wished to cheat under the condition Rhine described could deceive the investigator” (Christopher 1970: 24-25). Rather than admit that when controls are tightened it becomes more difficult to be deceived, many psi researchers have concluded that the controls interfere with psychic power by destroying the trust that is necessary for psychic powers to work. Critics consider this an ad hoc hypothesis.
But, as we noted above, we cannot justifiably assume that all the Pearces in the psychic world have cheated. So, what other explanation, besides actually having psychic powers, could explain the ability of some people to do significantly better than chance in guessing experiments?
Before attempting to offer some alternative explanations, I should first note that there is no way I, or anyone else, could know why each and every individual who performs better than chance in a given experiment did so. I should also note that I take it for granted that if a person has psychic abilities then that person should do significantly better than chance in guessing exercises. I should also note clairvoyance or telepathy may be the reason why some people do significantly better than chance in guessing experiments. And psychokinesis may be the reason why some people are able to appear to affect the output of random number generators (RNGs). Finally, I should note that if every possible explanation except psi has been shown to be false or highly improbable, it would be reasonable to conclude that psi is the best explanation for results that deviate significantly from chance in guessing or RNG experiments.
other possible explanations
Here are just a few possible explanations for data indicating significantly greater than chance results in psi experiments: selective reporting, poor experimental design, inadequate number of individuals in the study, inadequate number of trials in the experiment, inadequate number of experiments (e.g., drawing strong conclusions from single studies), file-drawer effect (for meta-studies), deliberate fraud, errors in calibration, inadequate randomization procedures, software errors, and various kinds of statistical errors. If any of the above occur, it is possible that the data would indicate performance at significantly greater than expected by chance and would make it appear as if there had been a transfer of information when there had not been any such transfer. It is also possible that information is being transferred, but not telepathically, through sensory leakage. Or, maybe some people have an ability to subconsciously recognize hidden patterns.
There are other scientific approaches to guessing experiments that don’t presume deviation from chance implies something paranormal. For example, the research of Peter Brugger and Kirsten Taylor (2003) looks at guessing experiments from a neuroscientific perspective. Brugger, a neuropsychologist at University Hospital, Zurich, and Taylor, a postdoc in experimental psychology at Cambridge University, have argued that the data in many ESP experiments that have found some people do significantly better than chance at guessing such things as card or die faces in allegedly randomized trials may indicate that some people have an ability to subconsciously recognize hidden patterns. Such ability would not require anything paranormal to explain. Brugger has received a grant to study “implicit sequence learning,” as this alleged subconscious ability is called. The Cogito Foundation, which is funding the study, has required that a parapsychologist also be involved. John Palmer, current research director of the Rhine Research Center, will be spending about a year in Zurich to conduct the research with Brugger. (Personal correspondence.)
Brugger and Taylor don't explore this, but it is possible that implicit sequence learning explains why Rupert Sheldrake was able to get statistically significant results in his study on staring that showed that some people can tell when others are staring at them. According to Marks and Colwell (2000), the study was flawed because it used a "random" process that had a distinct pattern. The results were replicated only when the same random process was used. But when a truly random process was used, Sheldrake’s results couldn't be replicated by Marks and Colwell.
Brugger and Taylor suggest that the letters ESP might better refer to Effect of Subjective Probability than to extrasensory perception (2003). In other words, they propose that the data from guessing studies might really indicate that some people have an ability that others don’t, but that this ability involves processes understandable by current scientific theories and knowledge.
Scientists like Peter Brugger are trying to find physiological bases for alleged paranormal experiences. For example, he and his associates have published reports on studies that examine a physical basis for such experiences as hauntings, the out-of-body experience, the feeling of a presence, the doppelganger experience, and phantom limb sensations. One of the experiments that he and Palmer plan to do will involve giving an experimental group L-dopa, which will increase levels of dopamine in the brain. They want to investigate the influence of dopamine on implicit sequence learning, guessing accuracy, and the reactions to feedback. Brugger’s research indicates that “Dopamine seems to help people see patterns,” (2002: New Scientist). People with high levels of dopamine are more likely to find significance in coincidences and pick out meaning and patterns where there are none. He thinks that brain chemistry might account for many paranormal experiences even in the absence of external stimuli.
There are other alternative explanations, as well, though some of them seem farfetched. For example, it is possible that statistically significant deviation from chance in psi experiments is caused by Zeus, aliens, angels, ghosts, Jehovah, jinn, or any one of a number of beings who dwell in other dimensions. These beings may be playing with parapsychologists, as James Alcock suggested with the Zeus hypothesis. Or they may be unwitting conduits of data transfer. Perhaps dolphins are picking up information telepathically from aliens and relaying it to subjects in psi experiments. As I said, some of these alternative notions are pretty farfetched, but they are possible nonetheless, and, in my view, just as viable as the psi hypothesis.
We should also note that the notion of statistical significance itself is an arbitrary concept and carries with it no necessary connection with our ordinary notion of importance. Statistical significance only tells us the probability that a given statistic is not spurious or due to a statistical accident. Statisticians express the likelihood that a statistic is due to pure accident by referring to its P-value. For example, P<0.01 means that there is a one percent chance the stat is accidental. The most commonly used P-value in the social sciences and medical studies is P<0.05, where there is a one in twenty chance that the result was a statistical fluke. This standard can be traced back to the 1930s and R. A. Fisher. At that time, the number of data points that might be produced by a scientific study would have been counted in the hundreds, thousands, or tens of thousands. Today, some psi studies have more than ten million data points. Should we assume that a statistical formula that was developed rather arbitrarily for studies with much smaller quantities of data can be applied without modification for studies with millions of data points?
Parapsychology is not alone in worshipping at the altar of P<0.05, but it is the only science that concerns us here. A good example of mistaking statistical significance for actual importance was provided by Dean Radin and Roger Nelson in their assessment of the data collected by Robert Jahn, Nelson, and Brenda Dunne in the PEAR experiments on psychokinesis. The experiments consisted of subjects who tried to use their minds to affect machines. In over 14 million trials by 33 subjects over a seven-year period they found that their subjects performed at the 50.02% level when 50.00% was expected by chance. With such a large number of trials, this data plugs into some statistical significance formula and spits out the result that the odds against this happening by accident were beyond a trillion to one (Radin 1997: 140). Why am I not impressed?
So, when confronted with data that indicate subjects in psi experiments are performing at levels significantly greater than chance, why should we conclude that psi is at work? We shouldn't, unless we can exclude all other possibilities. Of course, we can never do this with absolute certainly. But unless we can demonstrate that it is highly probable that all other possibilities are false, we are not justified in concluding that psi is the correct explanation for the data.
Some of the possibilities can be excluded on the grounds that they are too farfetched to be taken seriously. For example, the likelihood that Zeus, Jehova, dolphins, angels, ghosts, jinni, or aliens are the cause of the data strikes me as rather beyond belief. However, what I consider to be beyond belief should be irrelevant to the testing of any hypothesis. Therefore, even these wild notions should probably be considered by the parapsychologist if he or she is to be thorough in the investigation.
Other possibilities don't strike me as farfetched simply because we have ample evidence that they have occurred on numbers of occasions. We have numerous examples of cheating by subjects, fraud by scientific investigators, sloppy controls, inadequate protocols, poor record keeping, file-drawer effects, drawing grand conclusions from single or small studies, misusing statistics, and the like. Furthermore, these examples are not limited to parapsychologists, but can be found in all the sciences.
How many clear, decisive, unambiguous examples of psychic ability do we have? So far, we have none.
Hence, it seems that there is little or no justification for assuming that deviation from chance in a psi experiment is evidence of anything anomalous or paranormal.
See also clairaudience, clairvoyance, decline effect, displacement, ESP, experimenter effect, ganzfeld, Uri Geller, magical thinking, medium, Raymond Moody, Nostradamus, optional starting and stopping, parapsychology, past life regression, precognition, psi-missing, psychic, psychic detective, psychic surgery, psychokinesis, psychometry, James Randi Foundation psychic challenge, remote viewing, retrocognition, séance, Charles Tart, telekinesis, and telepathy.
Palmer, John (1983). "In Defense of Parapsychology: A Reply to James E. Alcock," Zetetic Scholar, No 11, 39-70.