Copyright 1974, 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. 38, p. 47-55)

ABSTRACT: The objective of this series of tests was to compare the action of PK on two electronic binary random generators- one simple and the other complex- under psychologically equivalent conditions. In the testing room the subject pressed a switch to initiate each trial and he was given feedback by way of two different colored lamps as indicators of hits and misses.

The simple generator produced a one-step binary output as the basis for the subject to exert his PK on each trial. The complex generator, on the other hand, based the binary output for each trial on the majority decision of a rapid sequence of 100 binary events. The choice of which generator was active on each trial was randomly made and neither the experimenter nor the subject knew which generator was operating during a given trial.

Significant results were obtained with both generators, and there appeared to be no significant difference in the scoring between the two.

PK tests with dice, electronic equipment, and various other devices appear to have produced scoring of comparable magnitude regardless of the nature of the testing device. This raises the question of whether or not the outcome of a PK test is independent of the structure of the randomizer and the nature of the underlying random process.

The following study has been designed to test this question by comparing the action of PK on two random generators of different degrees of complexity. For this comparison the psychological test conditions were kept constant because a psychological preference on the part of the subject for working with one or the other generator could easily mask any effect due to the physical differences between the generators. The experiment was carried out during February and March, 1973, at the Institute for Parapsychology.


The apparatus was located in two rooms on different floors of the Institute. The electronic equipment was on the lower floor; and the feedback apparatus for the subject was on the floor above.

The Two Binary Random Number Generators

The simple generator (S) was identical to the one discussed in a previous report (Schmidt, 1970). It generated sequences of binary random targets, "heads" and "tails." The choice of target depended on a high-frequency switch which oscillated between two positions and was halted at a time determined by the random arrival of a radioactive strontium-90 decay particle at a Geiger counter.

The complex generator (C) was built from a high-speed random number generator (RNG) used in another, previous experiment (Schmidt, 1973). If the high-speed generator was triggered, it produced a string of 100 binary random numbers, designated as +1's and -1's, within three seconds. Each individual binary number was generated by a process similar to the one used in the simple generator. (The only difference was that for practical reasons the random time element was not furnished by radioactive decay but by the random output of an electronic noise generator. The high-frequency switch was stopped when the noise amplitude passed a certain threshold.)

In the present experiment the individual random numbers in a string of 100 numbers provided by C were not recorded. The only information that was recorded was the "majority vote," that is, whether more +1's or more -1's occurred in the string. The majority-vote selection was made by an evaluator box connected to the RNG. The box had two outputs, a "head" output and a "tail" output, and, depending on whether more +1's or more -1's occurred in a string, one or the other output emitted a signal. In the case of a tie (50 +1's and 50 -1's) no signal was emitted. Thus the complex generator made a binary decision (except in the rare case of a tie) by taking the majority vote of 100 individual random events occurring during a three-second time interval.

Previous to, and between, the PK tests with human subjects the randomness of the generators was tested experimentally. For this purpose the equipment was activated automatically at the rate of one trial every eight seconds (similar to the typical operation rate in the PK tests) while the output signals of the two generators were counted. In five such test runs of approximately 16 hours each a total of 40,306 random numbers was produced by each generator. Neither the total scores (average scoring rates 49.96% and 49.8% with S and C, respectively) nor the chi-square values for the sessions as units suggested any deviations from ideal randomness.

During the experiment the subject, in the upper room, was seated in an arm chair looking at two feedback lamps, one of which lit up when he made a hit, and the other when he made a miss. The lamps were mounted in two sockets, one marked "head" and the other, "tail". These sockets were attached to long cables so that the subject had the freedom to arrange the lamps on a table in front of him or to attach them to hooks on the wall. As a safeguard against generator bias for either side, the test procedure was so arranged that in approximately half of each session the target was the lamp in the "head" socket, and in the other half, the lamp in the "tail" socket. Four different colored lamps were available, and the subject was free to insert his favorite colored lamp into the target socket and to change the colors of the lamps as often as he pleased. He was also provided with a switch at the end of a cable long enough to reach his chair. The subject pressed the switch to activate the test equipment for each trial. His task was to try to make the target lamp light up each time.

When the subject triggered his hand switch, both generators were activated. Only one was chosen, however, to be linked up to the feedback display for that trial. This choice was made by a selector switch whose position was determined by a prerecorded binary random number sequence stored in a tape reader. After each trial the tape was advanced by one step, and the next number on the tape (+1 or -1) determined the position of the selector switch.

In order to have sensorially identical test conditions for each trial, there was a three-second delay in the trigger line to the simple generator so that both generators would take the same length of time (approximately three seconds) for producing a binary output after the subject had triggered his switch. Neither the experimenter nor the subject knew which generator was momentarily in action.

An ink pen-recorder on the lower floor registered the following data for each trial:

a. Which generator was the active one, S or C.
b. The output of the active generator ("head" or "tail").
c. The output of S for the trials when S was not the active generator.
d. The target side.

Two electromechanical counters on the table in front of the subject in the test room indicated the total cumulative score. The tape record showed how the hits were distributed among the two generators. The recording (c) of the output of S even when S was not the active generator (and the subject was thus not motivated to obtain a particular output) was made with regard to the possibility that a previous effort of a subject to obtain, say, a "head" on one generator might "carry over" into the next trial even if that generator were no longer the active one.

No systematic study was intended with regard to how the scoring rates might depend on the subjects' attitudes or on other psychological factors. The subjects were encouraged to relax and to visualize the target color in some pleasant, possibly emotionally charged, context, and to start a trial (by setting the switch) only when they felt as completely as possible "immersed" in the target color. No attempt was made, however, to enforce this approach. Thus, some subjects would even get tense and try to "enforce" the lighting of the target lamp.

The subjects made their contribution to the test in one or in several sessions. Each session was continued only as long as the subject was optimistic and interested and felt able to concentrate well on the target color. Sessions were frequently interrupted by coffee breaks, walks on a porch, or by informal conversation. Much time was devoted to the discussion of how one might best actively visualize the target color.

Objectives of the Experiment

The questions to be studied in the experiment were these:

1. Is there evidence of PK under both conditions; i.e., when the simple RNG is active and when the complex RNG is active?
2. Is there an indication of a general scoring difference in the sense that, for example, one of the generators is "easier" to influence by PK?
3. Do some subjects score significantly differently under the two conditions?


After some encouraging exploratory studies a pilot test comprising approximately 1,000 trials was made.

The subjects in the pilot test were members of the Institute and others who had obtained promising PK scores in previous tests with different equipment. These subjects were not pretested; the scores were counted from the start. Only four subjects participated in the pilot. One of them was the experimenter, subject H in Table 2.


Table 1 shows the total score to be significant (CR = 3.0; P = .002, two-tailed). The main contribution to this positive score came from trials with the simple generator, which yielded a CR of 3.7 (P=.0002, two-tailed). The difference between the scoring on S and C was suggestive (CR=2.2; P=.02, two-tailed). The trials on S when it was inactive were at the chance level.

Table 2 gives the results for the individual subjects.

Table 1


Simple Generator ActiveComplex Generator Active Total Active TrialsTrials with Simple Generator Inactive
No. of trials5154961,011496
Scoring rate58%51.2%54.7%50%
*P = .002 (two-tailed).
**P = .0002 (two-tailed).

Table 2


N_s = Active trials with simple generator.
N_c = Active trials with complex generator.
D_s = Hits-misses for active trials on complex generator.
D_c = Hits-misses for active trials on simple generator.
D_s*= Hits-misses for nonactive trials of simple generator.


Since the pilot test gave significant evidence of PK, a confirmatory test containing three sections each was planned and completed. In sections 1 and 2 the subjects were members of the Institute and others who had had good PK scores in previous tests. In section 1 they knew that the experiment aimed at a comparison between the two generators; in sections 2 and 3 no mention was made of this comparison. Except in two cases, the subjects in sections 2 and 3 were told only that the experiment was aimed at learning somewhat more than just the existence of PK and that the experimenter was interested in the effect of distance, the nature of the generator, and the usefulness of vivid visualization in PK tests.

The subjects for section 3 were selected from random visitors to the Institute on the basis of informal tests of about 10 trials with the experimental equipment. The only difference was that in these selection tests the simple and complex generators were not alternated. The simple generator was used continuously. Subjects who showed a positive score and expressed interest in this particular test were selected to take part.

The number of trials to be made in each section was preset at 1,000. (Table 4 shows that actually about 10% more trials were collected. These additional trials were included in the evaluation, but the results are practically unchanged if only the first 1,000 trials in each section are considered.) The number of trials for each subject was specified to be approximately 200 trials for each of the 5 subjects in section 1, 100 trials for each of the 10 subjects in section 2, and 50 trials for each of the 20 subjects in section 3.

During all these tests the experimenter was in the same room with the subject but he did not look at the target lamps.


The overall scores in the confirmatory experiment with both types of generators were positive and statistically significant. (See Table 3.) The 1,695 trials when S was active gave a CR of 4.40 (P< 10^-5, one-tailed).

Table 3


Simple Generator ActiveComplex Generator Active Total Active TrialsTrials with Simple Generator Inactive
No. of trials1,6951,6093,304 1,609
Scoring rate55.3%53.8%54.5%50.7%
*P ~ 10^-3 (one-tailed).
**P < 10^-5 (one-tailed).
***P ~ 10^-7 (one-tailed).

Table 4


SectionSimple Generator Active Complex Generator ActiveTotal Active Trials Trials with Simple Generator Inactive
No. of trials15725471,119547
Scoring Rate155.8%53.8%54.8%52.1%

The 1,609 trials when C was active gave a CR of 3.0 (P ~ 10^-3, one-tailed). The difference between C and S was not significant, although the scores on S in all three sections were slightly higher than the scores on C. (See Table 4, which gives the results of the individual sections.) The scores on S, when it was inactive, were insignificant; i.e., there was no evidence of a "carry-over" effect.

Table 5 gives the scores for the individual subjects in each of the 35 sessions. To discover whether or not some of these subjects might have differentiated their scoring between the two generators, the 35 individual CR_d values were calculated and gave

Sum(i=1...35) (CR_d^i)^2 = 30.5

which, with 35 degrees of freedom, was not significant. Two further tests, the sign test and the Wilcoxon test, also failed to indicate any individual scoring differences between the two generators.

Table 5


Section 1


Section 2


Section 3

Grand Total1,6951,6971211,60929


The experiments compared the PK action on two binary random generators of different structures. During the experiments the two generators were automatically interchanged so that neither subject nor experimenter knew which generator was momentarily in action.

Under these conditions PK effects on both generators were observed, but there was no significant difference between the two scoring rates.

Whereas the sensory conditions for working with the two generators were identical, one cannot completely rule out the possibility that either the experimenter or the subject could have parapsychically known which generator was momentarily the active one or that the experimenter could have applied some goal-oriented PK so as to obtain some particular result when he later compared the two scoring rates. Thus the psychological equivalence of the two testing conditions would not rule out a parapsychological experimenter effect. Assuming tentatively that such psi-induced differences between the two conditions are small, the findings may stimulate a further study of the following "equivalence hypothesis":

If we have two systems which generate random signals, and if these signals are, in the absence of PK, statistically equivalent (i.e., indistinguishable), PK affects the two systems to the same degree provided we use sensorially equivalent conditions.


SCHMIDT, H. A PK test with electronic equipment. Journal of Parapsychology, 1970, 34, 175-81.
SCHMIDT, H. PK tests with a high-speed random number generator. Journal of Parapsychology, 1973, 37, 105-18.

Institute for Parapsychology
College Station
Durham, N.C. 27708

RetroPsychoKinesis Project Home