Copyright 1993, Journal of Parapsychology. Reproduced with permission. To purchase a complete issue write Journal of Parapsychology, 402 N. Buchanan Blvd., Durham, NC 27701, USA ( For copies of complete articles, write The Genuine Article, ISI, 3501 Market St., Philadelphia, PA 19104 USA.



(Originally published in Journal of Parapsychology, Vol. 57, Dec. 1993)

ABSTRACT: The author summarizes five experiments in which he studied the psychokinetic (PK) effect (the mental influence on the outcome of chance processes) under tight supervision by independent observers. Through the use of prerecorded random events as targets, the observers could evaluate the results independently, without having to trust the reliability of the author or his equipment. The total of these five studies, which represent all the work done under external supervision, produced an effect deviating by 3.67 standard deviations from chance expectancy. The odds against such an outcome are about 8,000 to 1. Thus, the results support the extstence of a PK effect on prerecorded random events, in agreement with previous experiments. The observed PK effect is inconsistent with current quantum theory. It shows that the theory is not correct when applied to systems that include human subjects. Furthermore, the existence of a weak mental effect on the outcome of chance events cautions the physicist to be careful in the interpretation of results that are based on relatively few chance events.

A psychokinetic (PK) effect - a mental influence on the outcome of chance processes - was first reported by Louisa and J.B. Rhine (Rhine & Rhine, 1943) from experiments with dice. Through the introduction of electronic random number generators based on quantum randomness (Schmidt, 1971), the experiments became more easily accessible and even more challenging to physicists. Today, a large number of successful PK experiments with random number generators have been reported (Radin and Nelson, 1989). It has not yet been possible, however, to stabilize and strengthen the statistically weak effects so that they can be easily demonstrated on demand.

The discovery of PK effects on prerecorded random events (Schmidt, 1976) did not make the gathering of data less arduous, but it permitted the inclusion of independent observers who, with little investment of time and effort, could obtain first-hand evidence for any PK effects that might occur. All five experiments to be discussed here included such independent observers.

In a typical PK experiment, a random number generator produces a binary random sequence that the subject tries to bias in a certain manner. Let us take, for example, the case where the random generator produces in each test run a sequence of 100 binary events displayed as a sequence of 100 red (for a 0-bit) and green (for a 1-bit) light flashes, while the subject is instructed to mentally enforce the appearance of more red than green flashes. The sequence of red and green signals is stored on floppy disk, and a score that is measured by the difference between the numbers of red and green signals is displayed and recorded at the end of the run (see Figure 1).

Figure 1 here

Figure 1. PK experiment with random events

The arrangement for an experiment with prerecorded random events (see Figure 2) is similar, but there is a time delay between the generation of the random events and their display to the subject.

Figure 2 here

Figure 2. PK experiment with prerecorded events

First, the random signals for a large number of test runs are generated and recorded on floppy disk. Days or months later, the stored signals are translated by the computer into a corresponding sequence of red and green light flashes while the subject tries to mentally enforce the appearance of more red than green signals.

Experiments have shown that selected subjects can succeed in this arrangement (Schmidt, 1976, 1987), provided that the prerecorded events have not been inspected by anybody before the subject makes the PK effort.

To let an independent observer participate in the experiment, we prerecord the bit sequences for all test runs planned for the experiment, and we let a computer make a printout of the scores for all the test runs. Being careful that nobody can observe the scores, we send a sealed copy of the printout to the independent observer. Leaving the printout sealed, the independent observer randomly specifies for each test run whether the subject's PK effort should be aimed at a high score (excess of red signals in the example) or a low score (excess of green signals).

In the subsequent PK test sessions, the subject follows for each test run the independent observer's assignments, aiming for an excess of red or green signals, respectively.

At the end of the experiment, the independent observer opens the sealed printouts. If the subject's effort was successful, the independent observer can confirm this first-hand by a tendency of the scores to point in the directions (positive or negative) that he had randomly specified. Because the independent observer had randomly assigned the directions, he or she can be certain that no such systematic tendency should occur (in the absence of the claimed anomaly). Using an appropriate statistical method, the independent observer can assess the significance of such a tendency without having to consider the reliability of the experimenter and his procedures.

Test Arrangement with Distant Subjects

The five experiments under discussion followed this general outline with minor variations. The most notable difference was the use of a variety of different feedback options and a test arrangement permitting subjects to work at home at their convenience instead of coming to the laboratory.

In the first experiment, I mailed to the participants small computers that could be connected to TV sets. During a run of typically one minute, the subject saw a pendulum swing on the screen, with randomly varying amplitudes. The amplitude variations were dictated by a predetermined binary random sequence of zeros and ones. After each half-cycle of the swinging pendulum, the computer read the next bit and, for a 1 or a 0, respectively, increased or decreased the amplitude (if possible within the available range). Depending on the target assignment, the subject tried either to make the pendulum swing with maximal amplitude over the whole screen, or to keep the swinging confined to a narrow region at the center of the screen.

The last two experiments provided a similar display, without the use of a TV screen. The test machine showed a row of 27 lamps on which the subject saw a light swinging back and forth. The system was controlled by a microprocessor chip that, in addition to the test program, held the prerecorded random events that determined the swing pattern.

Most of the subjects were not aware that this was a PK test with prerecorded events, because it looked like an ordinary PK test in which the subject tries to affect the display in a certain manner and receives a success score at the end of a run.

In Experiments 2 and 3, the subjects received only auditory feedback. In a typical case, the subject's task was to extend or to shorten tone intervals of random duration, prerecorded on cassette tapes. These tapes were mailed to the subjects with appropriate instructions.

These experiments were particularly simple to set up because only cassette tapes had to be mailed out instead of expensive microprocessor-controlled test machines. There was a possible disadvantage, however, insofar as many subjects noticed that this was not a standard PK test. They wondered how they could affect the tones, which had already been recorded on the tape, and I doubt whether the explanations I was able to provide gave them much comfort.

More information about the forms of feedback used, as well as other details, is provided in the Appendix.

Proposed Interpretations of PK Effects on Prerecorded Data

To our naive intuition, PK effects under time displacement may appear particularly puzzling. With the random events already generated and recorded, we feel that the subject's mental effort comes too late to have an effect. Let us remember, however, that our naive intuition is equally at a loss in explaining better known psi phenomena such as clairvoyance and precognition.

One logically consistent viewpoint is that the mental effort of the subject in the test session has a retroactive effect on the moment the random events were generated (Schmidt, 1975, 1978). One might want random to say that the random generator "senses" that a subject will later make a PK effort, and behaves accordingly. A similar process, with the time order between cause and effect inverted, may also be found in precognition.

Another viewpoint (Schmidt, 1982, 1984) is based on the idea of quantum theory that only those happenings that have been observed are physically real. In this case, the random events recorded on disk, as well as the corresponding printed scores, are not yet real; nature has not yet decided on the outcome until somebody has looked at the outcome. Thus it is only at the moment of observation by the PK subject that nature decides for "red" or "green." Then, the subject's mental efrort does not have to reach into the past, because the "collapse of the state vector," nature's final decision for one of several possible branches of reality, occurs during the test session.

The Effect of Preinspection of the Prerecorded Data

From the viewpoint just mentioned, the PK effects should disappear if somebody has looked at the prerecorded scores previous to the PK test session. In this scenario, the preinspection would collapse the state vector so that the subject would find nature in a decided state, with no opening left for PK to act. For this reason, the initial PK experiments with prerecorded events precluded such preinspection. On the other hand, experiments with preinspected prerecorded data are most interesting because they might discriminate between the two mentioned viewpoints.

Two previous experiments were concerned with the effects of preinspection. In one experiment (Schmidt, 1985) where the preinspection was very thorough (letting the preinspector experience the scores as vividly as they were later experienced by the PK subject), the preinspection inhibited the subsequent PK effort. (This result should be taken with some caution until it has been confirmed by other investigators.) Another group of experiments has indicated that the preinspection of the prerecorded data does not inhibit the subsequent PK effort if the preinspection does not provide immediate information about the resulting scores. These experiments (Schmidt, 1981, and the present Experiment 1) worked with prerecorded and preinspected random seed numbers as follows.

Use of Prerecorded Random Seed Numbers in the First Experiment

In the arrangements described so far, the random bits on which the subjects directed their PK efforts were individually generated and stored. In the first experiment, however, a true random number generator was used only to generate, for each test run, one 19-bit seed number. A quasi-random algorithm then derived from this seed number the larger number of bits required for the test run. One might feel that the reduced amount of randomness, entering only through the random seed number, gave the subject fewer chances to succeed. On the other hand, there are theoretical models that predict the same success rate as in the earlier arrangement (Schmidt, 1975, 1978). Furthermore, previous studies have shown that, indeed, PK tests with prerecorded seed numbers can produce PK effects (Schmidt, 1981), and that the PK effects even persist when the seed numbers have been observed before the PK session. Note that the observer of a seed number receives no direct information on the resulting score, unless he works through the complex algorithm that derives the score from the seed number.

At the start of the first experiment, the experimenter generated and preinspected lists of random 6-digit seed numbers. The independent observers obtained an open printout of these seed numbers, as well as the information about how to derive the scores from the seed numbers at the final stage of the experiment. Thus, the open list of seed numbers was equivalent to the sealed list of scores provided in the other experiments. The independent observers gave a random target assignment to each printed seed number, requesting a high or low resulting score, and returned the assignments to the experimenter.

A practical advantage of this arrangement was that no prerecorded data had to be stored in the test device. Instead, at the start of each test run, the subject typed the next seed number from the prepared list of random numbers into the test computer. From this list, the computer calculated and consecutively displayed the resulting binary sequence, while the subject made a mental effort in the specified direction; and, at the end, the computer displayed the final score.

At the end of the whole experiment, the independent observers typed the seed numbers into an initially supplied test computer (or used a computer algorithm, also supplied) to derive the scores from the seed numbers and check the results.

The positive results obtained under these conditions need not invalidate the viewpoint of our state-vector-collapse model, but they restrict the kind of observations that do collapse the state vector and force nature to decide between different possible branches of reality.

Statistical Design of the Five Experiments

The five experiments used different subject populations, different psychological approaches, and a variety of feedback displays. This made the experiments more interesting for experimenter and subjects. On the other hand, the overriding goal of all experiments was to present evidence of psi effects to the independent observers; and for the independent observers much of this variety was irrelevant and not even visible.

A certain inconvenience for the experimenter is the need to specify the length of an experiment in advance in order to avoid the well-known problems connected with "optional stopping." In the present experiments this inconvenience was alleviated by the following method.

Each of the five experiments was subdivided into a prespecified number (n) of "units" (see Table 1), where each unit represented a mini-experiment that could be evaluated (by a prespecified method) in terms of a z value measuring the deviation of the result from chance in units of one standard deviation, with positive values indicating a deviation in the target direction.



ExperimentUnits (n)Observers
110Morris, Rudolph
38Morris, Hardin

From the z values of the individual units z1, z2,...,zn, a final z value for the experiment was derived as

z = (z1 + z2 + ... + zn) / sqrt(n).

Whereas the number (n) of units of the whole experiment had to be specified at the start, the length and other details about a unit had to be specified (and to be communicated to the independent observers) only at the start of this unit. This gave the experimenter some flexibility in adjusting the length of the next unit to fit the available subjects and the available time.

At the start of a unit, the experimenter and independent observer agreed on the evaluation method, and then the independent observer received the score printout, which also specified the length of the unit.

The independent observer, in turn, sent the target assignments to the experimenter, and the test sessions could begin. At the end, the experimenter and the independent observer separately evaluated the results from the unit. (Details about the evaluation methods are given in the section on Evaluation of the Different Units.)

Differences in the Feedback and in the Score Definition

For different kinds of feedback, providing different PK tasks, the score of a run was defined such that this score reasonably reflected the subject's success in the PK effort. Let us discuss this for the three major classes of feedback that were used in the experiments:

1. One-dimensional unrestricted random walk. Consider a long linear string of lamps with the center lamp lighted at the start. A binary-bit sequence moves the light, one step at a time, to the right for a 1 and to the left for a 0. The subject's task is to move the light as far as possible to the right. Then a score (for the independent observer's printout) can be defined in the same manner as in the example previously given, as the difference between the numbers of 1s and 0s in the sequence. Because the length of the lamp string is limited, the light is reset to the center whenever it reaches a side.

2. One-dimensional random walk with two boundaries. Consider as an example a linear string of seven lamps. Starting with the center lamp lighted, each 1 or 0 moves the light one step to the right or left, respectively, whereas the light stays stationary when the move would push it beyond the seven-lamp range. The subject's task is to move the light to the specified side and to keep it there as much as possible.

Let the positions of the lamps from left to right be given by x = -3,-2,...,+3; and let x(n) be the light position after the nth step. Then we can define the score, proportional to the average location of the light, as

Score = x(l) + x(2) + x(N),

where N is the total number of steps.

Many of the units used a slight modification of this display, in which the subject saw a pendulum swing with seven different amplitudes. After each half-cycle the amplitude A = x(n) + 3 could increase or decrease by one step, subject to the upper and lower limits. The subject's aim was an average high or low amplitude.

3. Random time intervals. An interesting task for the subject is the extension or the shortening of tones of random duration. To digitally produce a random time interval, one can generate random numbers in the range from 0 to (M - 1) at a regular rate and terminate the time interval after a 0 is generated. The probability for an interval length of m steps (m = 1,2,...) is given by

P(M)=pq^(M-1) with p=1/M, q= 1-P.

For reasonably large M values, (for example, M = 32), the time intervals appear practically continuously variable, with the same statistics as the time intervals between signals from a Geiger counter exposed to a weak radioactive source.

If a subject tries to extend the durations of n time intervals in a test run, the score can be defined as

Score = L(1) + ... + L(n),

where L(n) is the duration (in steps) of the nth interval.

Channeling the PK Effort in the Desired Direction

For units using the first two classes of feedback, the scores were printed into the sealed list for the independent observer, who in turn randomly assigned a target direction for each score. It was the experimenter's task to direct the subject's PK effort in the assigned direction, consistent with a high or low score, respectively. For this purpose, in principle, the experimenter could have provided the subject with the target list specifying the direction of the PK effort for each run. This was done in Unit(1,1), the first unit of the first experiment. However, the frequent switching in the target directions, with the need for mental readjustment, as well as the need to keep track of the target list, appeared undesirable to most subjects.

For the later units, therefore, the experimenter changed the original bit-sequences from which the scores were printed into a secondary bit-sequence for which the bits corresponding to runs with low-score assignment were inverted (Os and 1s interchanged). Note that an inversion of the bits in a run inverts the sign of the corresponding score. The secondary bit-sequence was stored in the test machine used by the subject, so that the secondary sequence determined the display during the run and the (secondary) score seen by the subject. In order to succeed, the subjects no longer needed to consult the assignment list. Rather, they could consistently aim for high (secondary) scores shown at the end of each run.

As an added convenience for the subject, the display during the run could be inverted by the flipping of a switch so that success indicated by high secondary scores could be associated, not only by motion to the right or high swing-amplitudes, but also by motion to the left or low swing-amplitudes. Thus the subject could freely set the ostensible aim that seemed most attractive for the moment.

For units using the third class of feedback, the situation was different. There, the lengths of the generated random intervals were prerecorded in the form of binary numbers, but the durations of the intervals (long or short) could not be inverted by inverting the bits in some binary sequence. Furthermore, it seemed psychologically important that the subjects should work consistently with the same target direction; for example, long intervals.

Therefore the experimenter set a fixed target direction at the start and let the independent observer's assignment enter as follows: Each test run was based on a set of prerecorded random numbers (specifying the lengths of the random intervals) with the score equal to the sum of these random numbers. For each test run, the experimenter prepared 10 such sets of numbers and inserted the corresponding scores into one line of the independent observer's printout. The independent observer then randomly decided which of the 10 entries in each line the PK effort should be directed on to so as to make this number larger than the other 9 "control data." Accordingly, the experimenter loaded the specified data set into the test machine which later displayed the data set (in the form of random time intervals) to the subject.

In order to evaluate the results, the independent observer simply rank-ordered the scores in each line and checked whether the selected scores had a tendency to exceed the other control scores.

Evaluation of the Different Units by the Independent Observers

For the majority of the units (except the units using random time intervals), the independent observers had specified the target assignments, that is, high or low values for the scores printed in the sealed list. The question was whether the values found after unsealing the lists tended in the specified directions. In Experiments 1 to 4, a rank order test was used to measure such a possible tendency and to translate it into a z value. This nonparametric test made no assumptions about the distribution of the data supplied by the experimenter (for details, see Schmidt, Morris, and Rudolph, 1986). Before the start of Experiment 5, the experimenter had found an equally valid but conceptually and practically simpler evaluation method, which was then used for Experiment 5, giving essentially the same result as the previously used method.

To explain the simpler method, consider, for example, a unit with 100 printed scores, S(1), S(2),... S(100). From this score sequence, a secondary sequence, S'(1), S'(2),...,S'(100), is derived by inverting the signs of all scores S(n) that received a "low score" assignment. PK success is now indicated by a tendency of the numbers S'(n) toward positive values. We base our evaluation on the sum

V= S'(1) + S'(2) + ... + S'(100).

For any given set of values S(n), because of the random inversion of the signs in going from S(n) to S'(n), the sum V is a random variable with expectation value V* = 0 and with the variance

(V*)^2 = S(1)^2 + S(2)^2 + ... + S(100)^2.

This is true under the null hypothesis, in the absence of PK effects. Because of the large number of contributing terms, the random variable V has near normal distribution, and we can measure the significance of a possible PK effect in terms of the deviation of V from chance in terms of one standard deviation z = V/sqrt((V*)^2).

The possible corrections suggested by deviations of V from a normal distribution become even less relevant at the end where several units are combined to form one final z value.

In the units using random time intervals, where the independent observer randomly selected one out of 10 numbers for the test score as opposed to 9 control scores, the question was whether the test scores were generally larger than the control scores. Accordingly, the selected score was compared with the other 9 scores in the same row. The selected score was assigned a rank r when r of the scores in the row were lower than the selected one. Thus, r can assume the integer values from 0 to 9. With the probability for ties negligible, the expectation value of r is 4.5, and the variance of r is also known (we are dealing with the statistics of a "ten-sided die"). The independent observer evaluated the total significance in terms of the r values for all rows of the printout. With expectation and variance of this sum known, a z value could be easily calculated.

The Independent Observers

In all five experiments, there was another person (sometimes two others) acting as independent observers who independently assigned the targets and calculated the results. Some of these independent observers were psi researchers, but we took careful formal precautions to guarantee that no single person, independent observer, or experimenter could have simulated high scores by fraud or human error.

The following list gives a brief description of the independent observers:

Experiment 1: Observers, Morris and Rudolph. Robert Morris presently holds the Arthur Koestler Chair for Parapsychology at the University of Edinburgh. At the time of the experiment, he headed a psi research group with the Department of Computer and Information Science at Syracuse University. Luther Rudolph was a professor at the School of Computer and Information Science at Syracuse University.

Experiment 2: Observer, Schlitz. Marilyn Schlitz holds a PhD in Anthropology, and presently works in the Psychology Department at Stanford University. At the time of the experiment, she was, like myself, a Research Associate at the Mind Science Foundation. She has an active interest in parapsychology and has published studies of her own.

Experiment 3: Observers, Morris and Hardin. Robert Morris (same as above) holds the Arthur Koesder Chair for Parapsychology at the University of Edinburgh. Larry Hardin was Professor of Philosophy at Syracuse University. He had no particular interest in parapsychology.

Experiment 4: Observer, Braud. William Braud is now Director of Research at the Institute of Transpersonal Psychology in Palo Alto. He was a colleague of mine at the Mind Science Foundation, and has worked extensively on PK effects acting on living systems (with their inherently random features) rather than on electronic random number generators.

Experiment 5: Observer, Stapp. Henry Stapp is a theoretical physicist at the Lawrence Berkeley Laboratory. His publications relate in particular to elementary particle physics and the foundations of quantum theory. He is interested in the role of consciousness in physics, but maintains a sceptical outlook toward parapsychology.

Random Number Generation

The experimenter used a combination of a true random number generator and a quasi-random algorithm to generate the primary random bit-sequences and the random time intervals needed for the experiment.

The random generator utilized the timing of radioactive decay as the basic source of randomness (Schmidt, 1970) to generate binary random bit-sequences. Regular randomness checks never indicated any malfunctions of this device. Nevertheless, as an additional precaution against the unlikely event of generator malfunction, the resulting bit-sequence was combined through the logical XOR operation with a cornputer-generated quasi-random bit-sequence, based on the multiplicative algorithm R'= R*M (mod p) with p = 2^19 - 1 = 524287, M = 242292. This procedure ensured that even a complete breakdown of the "true random generator" could not lead to a systematic bias in the final bit-sequence.

Random Target Assignments

The independent observers used a different random method to give target assignments to the scores printed in the sealed lists.

The inclusion of the independent observers strengthened the reliability of the experiment in two respects. First, even though the experimenter had taken precautions to guarantee the desired randomness, the presence of one more independent random source was an additional safeguard against malfunctions on the part of the experimenter. Second, the arrangement precluded fraud by the experimenter as well as fraud by the independent observers.

In Experiment 1, the independent observers used their own random number generator to determine the target assignments. Following a suggestion by Robert Morris, the assignments in the subsequent experiments were derived from future weather data. A prespecified number of days (for example, 7 days) after the independent observers received the sealed score printout, they bought a prespecified newspaper (for example, The New York Times) and derived from the last digit in a prespecified weather column a 6-digit seed number. For Experiments 2 to 4, this seed number served as an entry point into the RAND random number tables. The consecutive digits following the entry point determined the binary target assignments (high/low for even/odd digits), or the one of 10 scores in each line to be used as target in the case of random intervals.

In Experiment 5, at his own suggestion, the independent observer used a quasi-random algorid-im of his own choice rather than the RAND tables to derive the target assignments from the weather data.

Protection Against Fraud

To preclude fraud by the experimenter, it was sufficient that (a) the target assignments were determined after the experimenter had given the sealed score printouts to the independent observers, and (b) the experimenter had no further access to these records. In the two cases (Experiments 2 and 4) when colleagues from the same laboratory acted as observers, the observers kept the records securely in their homes rather than in the laboratory where the experimenter might gain access to replace or alter the records.

To preclude fraud by the independent observers, the arrangement of the first experiment called for two independent observers to watch each other. After the experimenter sent the sealed printout to the home address of one independent observer (Rudolph), the other independent observer (Morris) used his own random generator to print out the sequence of target assignments. Then the two independent observers met and exchanged copies of their records. A copy of the target assignment list was then mailed to the experimenter.

In Experiments 2 to 4, the target assignments, determined by future weather data and by the RAND tables, were beyond anybody's control, and the proper assignments calculated initially by the independent observers could later be verified independently by the experimenter. There was no room for fraud even with a single independent observer. Nevertheless, Experiment 3 used two independent observers. Indeed, the arrangement can easily accommodate any number of independent observers.

In Experiment 5, the situation was slightly different insofar as the single independent observer used a quasi-random algorithm of his own choice to take the place of the RAND tables. The experimenter did not know this algorithm at the start. This would, in principle, have enabled the independent observer to select an algorithm that would produce favorable scores. In view of the rather skeptical attitude of this independent observer, however, this possibility was considered to be negligible.

The Main Results

Table 2 lists the z values z(l),...,z(5) that determine the significance levels obtained in the five experiments. All experiments gave score deviations from chance in the desired direction, but the z values for most experiments were too low to provide independent statistical significance. Combining the results of all five experiments, however, we obtain a total z value of

z(tot) = [z(l) + z(2) +...+ z(5) ] /sqrt(5) = 3.67.

The odds that the observed score or a higher total score could result from chance are approximately 8,000 to 1.

Thus, the five experiments together provide convincing evidence for the existence of an anomaly. (Detailed information on the individual experiments and their units is provided in the Appendix.)

The reasons for compiling the results of the five experiments at this stage are practical ones. First, such compilations seem appropriate at some stages, and second, the author lost his financial support after the fifth experiment, so that similar experiments might not be expected in the near future.



12.71Schmidt, Morris, and Rudolph (1986)
21.66Schmidt & Schlitz (1988)
30.62Schmidt, Morris, & Hardin (1990)
41.98Schmidt & Braud (1992)
51.23Schmidt & Stapp (1993)

Additional Questions Studied

Although the demonstration of the existence of PK under external supervision was the main objective, additional questions were explored in some of the experiments. In Experiment 2, for example, the performances of meditators and nonmeditators were compared. The results suggested that meditators performed better than nonmeditators. However, when only meditators were used in Experiment 3, the total score was disappointing, so that even the selection of meditators does not guarantee reliable performance.

Experiment 4 compared the PK action on prerecorded random events with PK action on momentarily generated random events. For this purpose, each test machine was equipped with an internal random number generator using electronic noise as the basic source of randomness. During a test run, the bits coming from this random generator were interlaced with prerecorded bits, and that bit mixture was displayed so that the subject could not distinguish between prerecorded and momentarily generated bits. The PK scores on the two types of bits were internally recorded, but only at the conclusion of the experiment were they inspected and evaluated. The result was surprising in that it showed at least a suggestive difference between positive scoring on the prerecorded events, (z = 1.98) and slight PK-missing (z = - 0.23) in the momentarily generated events. It is interesting to note that the experimenter was motivated to obtain high scores on the prerecorded events, which were monitored by the independent observer. The outcome of the directly generated events, which could not be monitored by the independent observer, however, was of little concern to the experimenter. We must leave the question open for the moment as to whether the observed difference was a fluke of chance, an experimenter effect, or some other of the "differential effects" that have puzzled researchers in the past.

Experiment 5 compared two kinds of binary events. For one kind, the corresponding score was preinspected before the subject made the PK effort. The two kinds of binary events were again interlaced so that the subject could not distinguish between them. And also the experimenter learned the possible scoring differences on the two kinds of events only at the end of the whole experiment. The outcome of the experiment was a slightly positive score on the not-inspected events (z = 1.23) and a slightly negative score (z = - 0.93) on the preinspected events. At the start of the experiment, it was considered as possible or even likely that preinspection might inhibit the PK effects. Therefore, it was decided to use as evidence for the existence of a PK effect only the results from the not-observed events. Thus, the experimenter was motivated to obtain positive scores on the not-inspected events, but was neutral toward the preinspected events. With the difference between the two scores not statistically significant, we have to wait for the outcome of further experiments.


Previous to this study, several independent researchers had reported the existence of a PK effect. The present study confirms the existence of the effect under particularly well-controlled conditions where the participation of independent observers precludes experimenter error, or even fraud.

The PK effect appears as an anomalous correlation between the outcome of random events and the mental state of a human subject observing the outcome. The effect is only partly under voluntary control and may depend on subconscious expectations, wishes, and fears of the observer.

Although the effect is usually weak, it may be of practical importance in cases where conclusions have to be drawn from limited statistical evidence. A physicist trying to confirm his own theory experimentally might subconsciously generate a PK effect that could shift the outcome by, say, two standard deviations in the desired direction. Similarly, the fear of failure might induce PK effects opposite to the desired direction. This could play a practical role in critical procedures such as the launching of a satellite where, as we know from experience, elements of chance cannot be completely excluded, and where even a small reduction of the failure rate could be economically significant.

The most challenging aspect of PK is its incompatibility with current quantum theory. The experiments indicate that the outcome of quantum jumps, which quantum theory attributes to nothing but chance, can be influenced by a person's mental effort. This implies that current quantum theory is wrong when experimentally applied to systems that include human subjects. It remains to be seen whether the quantum formalism can be modified to include psi effects, and perhaps even to clarify the still somewhat puzzling role of the human observer in the theory.

The use of prerecorded events in the present study served two purposes. First, it permitted the inclusion of independent observers, and, second, it emphasized the difference between PK and the known physical mechanism. Furthermore, the experiments raised a new question, the effect of preinspection of the prerecorded data. The future study of this question may shed new light on the psi mechanism and the role of consciousness.



Table 3 provides details on the individual units, The unit labels in column 1 give the number of the experiment followed by the unit number within the experiment. Each unit was divided into test runs, with a run lasting typically between 15 and 60 sec. Column 2 gives the number of runs in the unit and column 3 lists the approximate total test time in minutes. Comparing the test times for the units provides an approximate measure of the total time spent on the unit. In many cases, however, the subjects spent considerable additional time in preparing themselves for the test runs. In some cases (Experiment 2 and part of Experiment 3) the subjects received feedback through mailed-out cassette tapes. In these cases the numbers of tapes used, followed by a T, are given in column 2. Each tape held between 10 and 20 minutes of test time so that these units comprise a very large total test time.

The number of participating subjects in column 4 varied from one to over a hundred in some of the experiments with cassette tapes. The * next to the 1 in some units indicates that the author acted as the only subject. The author also participated in some of the other units of Experiments 1 and 5, mainly in an attempt to bring nearly completed units to an end.

The z values in column 5 were calculated independently by the experimenter and by the independent observer using the previously described procedures.

The different units used a large variety of feedback displays (column 6) which were provided either by small test machines that could be mailed or, in the case of mere auditory feedback, by cassette tapes. We will discuss here only the main features of the displays and ignore minor variations. The TV Pendulum Display used in parts of Experiments 1 and 3 was provided by a small microprocessor system that the subjects could connect to their own TV sets. The display showed a pendulum swinging on the screen, with the amplitude randomly varying on a 7-step scale (random walk with two boundary conditions). The subject had the choice (registered by the setting of a switch) of aiming either for a high or a low average swing amplitude. The task of increasing or decreasing a swing amplitude seemed psychologically quite captivating so that a similar display was also used in Experiment 4 (31 Lamps) and part of Experiment 5 (27 Lamps, I). In these cases the TV screen was replaced by a row of 31 or 27 lamps. With one lamp lighted at a time, the light performed a swinging motion around the center lamp, with randomly varying amplitudes. The PK goal was either to have the light swing with maximal amplitude over the whole range, or to remain with minimal swing amplitude near the center lamp. A frequently used variation was provided by a mirror symmetric display where the subject saw two lights "bouncing against each other" with randomly changing amplitudes.

An interesting auditory display (Clicks) was used in part of Experiment 1. The subjects listened through stereo headphones to two different click patterns. One pattern consisted of clicks sent simultaneously to both headphones. This was perceived as clicks in the center of the head. The other pattern consisted of clicks rapidly alternating between the headphones. This was perceived as clicks coming from the outside.



UnitRunsTest minutesNumber of Ss zDisplay
1/140401*1.66TV Pendulum
1/712012011.17TV Pendulum
2/1240T2400240-1.20 Sound Tape
3/16412811.47TV Pendulum
3/480T1600800.59Sound Tape
4/125625641.2131 Lamps
5/1200100100.3627 Lamps (II)
5/22001009-0.24" (II)
5/320020071.54" (I)
5/420020041.34" (I)
5/520020090.67" (I)
5/620010061.91" (II)
5/720010070.38" (II)
5/820020080.27" (I)
5/92002008-0.18" (I)
5/10200200120.64" (I)
5/11100501-0.64" (II)
5/12100501-1.3727 Lamps (II)
5/13100501*0.93" (II)
5/14100502-1.22" (II)
5/151005010.14" (II)
5/1610010010.83" (I)
5/1710010010.27" (I)
5/181001001*-0.61" (I)
5/1910010011.18" (I)
5/201001001-0.69" (I)

The display was driven by a binary sequence, with a 1 initiating the first pattern and a 0 initiating the second pattern. This display is promising for further work because the subject does not have to "reach out" in the PK effort, but has to affect things that are perceived as happening "inside the head."

A number of different sound displays provided by cassette tapes required the subjects to move tones up or down on certain scales (random walk with two boundaries), to extend the durations of tones, or to extend the duration of pleasant and shorten the duration of unpleasant tones (random time intervals).

In the cases discussed so far, the duration of a run was long enough (generally .5 min or more) so that the feedback could affect the subject's mental state, and with it possibly the scoring level.

In part of Experiment 5 (27 Lamps, II), the runs were very short, some only a few seconds, so that the subject had no time to reflect on the performance during the run, but, rather, approached the task with a "burst of energy" gathered before the start. (I had hoped that this approach would work particularly well with martial arts students, but the results were rather disappointing.) A typical test run displayed an unrestricted 128-step random walk of a light on a string of 27 lamps. The subjects tried to move the light toward a specified side, and whenever the light reached the right or left edge, it was automatically reset to the center to continue its random walk.

For further details, such as the selection of subjects, the reader is referred to the original reports.


RADIN, D., and NELSON, R. (1989). Evidence for consciousness-related anomalies in random physical systems. Foundations of Physics, 19, 1499-1514.
RHINE, L.E., and RHINE, J.B. (1943). The psychokinetic effect. Journal of Parapsychology, 7, 20-43.
SCHMIDT, H. (1970). Quantum mechanical random number generator. Journal of Applied Physics, 41, 462-468.
SCHMIDT, H. (1971, June). Mental influence on random events. New Scientist and Science Journal, 757-758.
SCHMIDT, H. (1975). Toward a mathematical theory of psi. Journal of the American Society for Psychical Research, 69, 301-319.
SCHMIDT, H. (1976). PK effect on prerecorded targets. Journal of the American Society for Psychical Research, 70, 267-291.
SCHMIDT, H. (1978). Can an effect precede its cause? Foundations of Physics, 8, 463-480.
SCHMIDT, H. (1981). PK effects with prerecorded and pre-inspected seed numbers. Journal of Parapsychology, 45, 87-98.
SCHMIDT, H. (1982). Collapse of the state vector and psychokinetic effect. Foundations of Physics, 12, 565-581.
SCHMIDT, H. (1984). Comparison of a teleological model with a quantum collapse model of psi. Journal of Parapsychology, 48, 261-276.
SCHMIDT, H. (1985). Addition effect for PK on prerecorded targets. Journal of Parapsychology, 49, 229-244.
SCHMIDT, H. (1987). The strange properties of psychokinesis. Journal of Scientific Exploration, 1, 103-118.
SCHMIDT, H., and BRAUD, W. (1992). New PK tests with an outside observer. Journal of Parapsychology, 57, 227-240.
SCHMIDT, H., MORRIS, R.L., and HARDIN, C.L. (1990, September). Channeling evidence for a psychokinetic effect to independent observers. (Mind Science Foundation Research Report.) San Antonio, TX: Mind Science Foundation. (See also abstract in Research in Parapsychology 1991.)
SCHMIDT, H., MORRIS, R.L., and RUDOLPH, L. (1986). Channeling evidence for a PK effect to independent observers. Journal of Parapsychology, 50, 1-15.
SCHMIDT, H., & SCHLITZ, M. (1988, December). A large-scale pilot PK experiment with prerecorded random events. (Mind Science Foundation Research Report.) San Antonio, TX: Mind Science Foundation. (See also abstract in Research in Parapsychology 1991.)
SCHMIDT, H. and STAPP, H. (1993, June). Study of PK with prerecorded random events and the effects of preobservation. Mind Science Foundation Research Report. San Antonio, TX: Mind Science Foundation (See also Journal of Parapsychology, current number.)

P.O. Box 296
Mora, NM 87732-0296

RetroPsychoKinesis Project Home