Printer Friendly

PK in a competitive computer game: a replication.

In an earlier paper in this Journal (Broughton & Perlstrom, 1986), we reported two PK experiments that were in the form of a competitive game of chance. We used the game format to take advantage of the motivation potential inherent in that approach (see Broughton & Perlstrom, 1986, for a discussion), and we added a competitive element to heighten certain aspects of the game's motivational character further.

The experiments involved a computerized dice game that could be played in a competitive mode, that is, with an opponent. It was based on a commercially available product called OINK! (Beagle Bros., San Diego, CA) and was heavily customized by us for use as a PK experiment. In brief: The participant sat in front of a video monitor (similar to a small TV) on which two die faces appeared each time the player pressed a button on a game paddle. The object of the game was to accumulate a higher score (summing the numbers represented by the two die faces) than one's opponent. The game proceeded in alternating turns of five "rolls" of the dice. Players were penalized for rolling doubles (one of the rules in the commercial product). The random numbers governing the die faces came from a true random source built into the computer.

Each participant played the game in two forms. In the noncompetitive form, the participant had the computer as an opponent, and this part of the procedure was portrayed either as a "warm up" or a "baseline" game. In the competitive form, the test program convincingly simulated a live opponent linked to the subject's computer by a telephone connection. Since most of the participants were students from nearby Duke University, we arranged to have our simulated opponents represent the University of North Carolina at Chapel Hill (a long-standing sports rival of Duke). Prior to the test session each participant completed the Sport Competition Anxiety Test (SCAT) (Martens, 1977), and immediately prior to playing the dice game (in either form) the participant completed a subset of Spielberger's State-Trait Anxiety Inventory (Spielberger, 1973) to measure state anxiety. At an earlier visit to the lab, each participant was asked to complete the Participant Information Form (PIF), a 10-page questionnaire developed at the Psychophysical Research Laboratories in Princeton, NJ.

The first of the two previously reported experiments (CGC1, meaning "Competitive Game of Chance No. 1") revealed no overall PK scoring or any effects of competition for the 50 participants. For those 23 participants who completed and returned the PIF, several interesting correlations emerged. Chief among these was a negative correlation between the participant's state anxiety immediately prior to playing the competitive game and the score obtained in the game: |r.sub.s~(21) = -.52, p |is less than~ .01, two-tailed. This suggested that the more anxious a participant was at the start of the competitive game, the lower he or she would score. Another negative correlation turned up between the scores of subjects who had practiced any mental discipline such as yoga (as indicated on a PIF question) and their competitive game scores: |r.sub.s~(21) = -.85, p = 3 x |10.sup.-7~, two-tailed. This finding suggested that participants who practiced a mental discipline were also likely to score poorly in the competitive game. The state anxiety prior to the competitive game and the practice of mental discipline were positively correlated. These correlations were not found in the data from the noncompetitive games.

That these results were obtained only with those participants who took the trouble to complete the lengthy PIF indicted to us that the findings might relate only to a subset of participants who were particularly interested in the research or were otherwise motivated to cooperate. Therefore, for a second experiment, which, owing to its automated nature, was largely completed before we had finished the analysis of the first, we predicted two findings: In the competitive game the scores of participants who returned PIFs would show a significant negative correlation with their state anxiety score just prior to that game, and their game scores would be negatively correlated with the practice of a mental discipline.

In the second experiment (CGC2), again there was neither an overall PK effect nor an effect of competition for all 36 participants. State anxiety scores in the competitive condition negatively and significantly correlated with game scores: |r.sub.s~(34) = -.35, p = .04, two tailed). For the 28 PIF returnees, the predicted negative correlation between scores and the practice of a mental discipline failed to appear. However, the predicted negative correlation between state anxiety and game scores was robustly confirmed: |r.sub.s~(26) = -.49, p = .01, one-tailed.

About a year after the second experiment, we conducted a third experiment (CGC3), essentially identical to, and intended as a close replication of, the second. Although the experiment was completed in 1986, various circumstances caused the data to remain unexamined until recently. We report here the results of that replication attempt.



A full description of the hardware and software can be found in Broughton and Perlstrom (1986). We shall only summarize the principal features here.

The experiment used an Apple II+ microcomputer with a color monitor, two disk drives, and a clock/calendar card. Standard "game paddles" (hand controllers) were used to register participant responses. All data were stored on diskettes.

Random numbers for the dice were obtained from an unmodified RNG board from the Research Institute for Parapsychology and Physics in Amsterdam. This device plugs directly into the computer and derives its randomness from twin Zener diode analog noise sources. Each random number was obtained according to the instruction X = 1 + INT (PEEK(50944)/32), with numbers greater than 6 discarded.

The psi-test software was an extensively modified version of the commercial game OINK!, a two-player dice game. In its psi-test form (renamed P-OINK), the hardware RNG supplied the random numbers, and the overall structure was changed to a fixed-length game consisting of 50 "rolls" of the dice in the form of 10 "turns" of five rolls. One of the rules of the original game was retained: Players had to avoid rolling doubles, which, if they did occur, had the effect of wiping out one's accumulated score within that turn. Routines were added to handle data storage including name, dates, times, and all rolls of the dice. One version of the game, used in the noncompetitive condition, recognized "computer" as the second player and automatically rolled the computer dice at a fixed pace. In the competitive version of the game, the program supplied a fake name for the competitor and displayed it with the cadence of a hunt-and-peck typist. The dice rolls for the simulated competitor were irregularly paced by superimposing a random element on a fixed pattern. The competitive version of the game was also preceded by a short program that simulated the establishing of a telephone link-up.(1)

The experiment had two checks for RNG adequacy. Each game incorporated an entire parallel control game: Every call to the RNG that produced a die face for the game was followed by a second call that was not displayed but was only stored as a control. The second check was performed by tests of the RNG at regular intervals during the conduct of the experiment. These tests consisted of collecting blocks of 100,000 random numbers 1-6 (using the same computer instructions as in the game), which were later analyzed by Good's Generalized Serial Test.

For security purposes, routines were incorporated to detect attempts to interrupt the program. The program also dated, time-stamped, and sequentially numbered all records so that any missing records would be noticed.

The feature of the game that caused scores to be wiped out when a double occurred meant that the theoretical mean and standard deviation had to be specially derived. For each turn (5 rolls), the mean chance expectation (MCE) for the score was 20.9343, which was multiplied by 10 (the number of turns in a game) to give the game MCE of 209.34. The standard deviation for a turn was 14.2496, and this was multiplied by |square root of n~ (where n = number of turns), giving 45.06 as the standard deviation for a game.(2)

Figure 1 shows one stage of the game as seen by the participant. Each time the participant pressed the button on the game paddle, two white squares appeared, and the "pips" of the die faces appeared sequentially, each accompanied by a short buzz on the speaker. Doubles were accompanied by an electronic "raspberry" and a message indicating how many points were lost. The game ended with enthusiastic beeps and whistles, color changes, and a flashing notice indicating who had beaten whom. The opponent's turns proceeded in the same manner, but each trial was initiated by the computer on an irregular schedule that gave the appearance of another player taking turns.

Psychological Instruments

In the preceding two experiments, we used the Sport Competition Anxiety Test (Martens, 1977) to assess general participant anxiety about competitive situations. As this proved to have little relationship with game scores, it was replaced by the trait part (form Y-2) of the Spielberger State-Trait Anxiety Inventory (STAI) (Spielberger, 1983). We shall refer to this test as the TAI.

A modified form of the complementary half of the STAI, the state part, was used to measure the participant's anxiety as he or she was about to start each game. This test (as used by Martens, 1977) consists of 10 questions with answers to be checked by the participant on a five-point scale. This will be referred to as the SAI.

We also used the Participant Information Form (PIF), a 60-item, 10-page questionnaire developed at the Psychophysical Research Laboratories in Princeton, NJ. This questionnaire includes a wide range of questions dealing with personal experiences and beliefs concerning psi, health habits, and other matters that may interest researchers.

General Procedure

Participants were recruited from the Institute's general participant pool of individuals who were already familiar with our lab through a prior visit. Most were Duke University students. PIF questionnaires were distributed at the participant's first visit to the lab, with instructions that it should be returned when he or she returned for an experimental session (not necessarily the P-OINK experiment). In general, most participants did return PIFs prior to beginning any experiments, though a few either forgot to bring it or had not gotten around to completing it on their next visit. For this experiment, only participants who completed a PIF form before doing the P-OINK experiment were considered as PIF returners.(3)

Participants were tested by appointment. The experimenter (J.R.P.) greeted subjects and administered the TAI. Afterwards, participants were taken to the testing area where the details and rules of the game were explained. Participants were instructed that they would each play two games: one was a "baseline" game, and the other was the "real" game consisting of a one-on-one match with a student from UNC.

The experiment proceeded at a relaxed pace, with the experimenter engaging in some minor role-playing to support the cover story. For each game, the experimenter escorted the participant into the testing room and started the game. Just before beginning, the participant was given the 10-question SAI, which took about half a minute to complete. The experimenter then took the form and retired to an adjacent area where he could monitor the game's progress by the sounds produced.

Following the second game, the participant was thanked and any questions were answered. Participants were not immediately debriefed concerning the simulation, lest the cover story be compromised. (The debriefing was done later by mail.)


The preset number of participants for CGC3 was 40. Forty-two subjects were drawn from the Institute's participant pool; one withdrew several days after the P-OINK session and another caused the internal security alarm in the program to be triggered. Although there was no cause for suspicion, it was thought best to replace this participant. The participants were 23 females and 17 males. They were tested from November 1985 through March 1986.

Hypothesis and Analyses

The results of the two preceding experiments led us to make one main prediction: There would be a significant negative relationship between SAI and competitive P-OINK scores for those participants who returned their PIF forms. This would be tested by Pearson's r with a one-tailed evaluation. (The prior experiments used Spearman's rho because this was one of several correlations, the rest of which used categorical data. Because the components of this single prediction are normally distributed, the Pearson test was more appropriate.)

The first experiment in the series had revealed a strong negative correlation between the practice of mental discipline and competitive P-OINK scores; therefore, we looked for the same trend here, though evaluation (by Spearman's rho) would be two-tailed.


For the 40 participants, participant game scores did not differ significantly from chance in either condition: competitive M = 214.2, SD = 37.6; noncompetitive M = 215.2, SD = 45.5). The "opponent" scores, which are also open to the participant's PK influence in his or her attempts to win the game, did not differ from chance in the noncompetitive condition: M = 211.0, SD = 47.1. However, in the competitive condition, opponent scores were significantly above chance: M = 230.8, SD = 39.0, | = 3.47, p = .001, two-tailed.(4)

Mean trait anxiety score for the group was 37.65. Mean state anxiety scores were 15.95 (SD = 5.03) for the noncompetitive condition, and 18.02 (SD = 5.41) for the competitive condition. Participant state anxiety immediately prior to playing the competitive game was suggestively higher than that for the noncompetitive game: t(38) = 1.78, p = .08.

Analysis of the SAI data for PIF returners (N = 35) revealed that the predicted relationship between SAI and participant game scores in the competitive condition was confirmed: r(33) = -.29, p = .048, one-tailed.(5)

Unexpectedly, there was an even stronger negative correlation between SAI and game scores in the noncompetitive condition: r(33) = -.42, p = .013.

Trait anxiety, which normally is strongly correlated with state anxiety, was correlated in this experiment also: competitive r(33) = .55, p = .001; noncompetitive r(33) = .57, p = .001. As with state anxiety, trait anxiety was negatively correlated with game scores in both conditions: competitive r(33) = -.34, p: .048; noncompetitive r(33) = -.33, p = .054, two-tailed.

The practice of a mental discipline as assessed by the PIF Question 33 ("Have you ever practiced any form of mental discipline, e.g., meditation, biofeedback, hypnosis, relaxation exercise? no, yes, consistently") was not correlated with participant game scores in the competitive condition: |r.sub.s~(33) = .00, n.s. In the noncompetitive condition a weak, nonsignificant negative correlation with game scores occurred: |r.sub.s~(33) = -.26, p = .13.

Control Data

A total of 18 general RNG tests were interspersed through the experiment. They were equivalent to 1,800,000 rolls of the computer die. Analysis up to the fifth order was done using Good's Generalized Serial Test on each test session. This yielded no unexpected departures from chance. Of the 90 resulting chi-square values, four exceeded the .05 level, and none exceeded the .01 level.

The parallel control data collected during the actual test sessions did not depart from chance expectancy on the game scores, and control game scores were not correlated with the true game scores. None of the significant correlations in the true data already reported were observed in the control data.(6)


Apart from the introduction of a randomized presentation of the competitive and noncompetitive conditions following the first experiment and the replacement of the SCAT with the TAI for the third experiment, the three experiments of this series were virtually identical. It is therefore appropriate to examine the three experiments in relation to one another and as a whole.

Prior to combining the data from the separate experiments for single analyses, we used an analysis of variance to test for differences between the experiments on the principal variables: subject score, "competitor" score, and SAI. This was done for all data as well as for PIF returners only. No significant differences were found.

The principal finding in this third series, present in all three experiments, was the negative relationship between the participants' scores in the competitive condition and the participants' self-rated level of anxiety immediately prior to playing. Figure 2 shows the correlations with 95 percent confidence intervals for the three experiments in each of the conditions for the subjects who returned PIF forms. Table 1 presents the numerical data.


 Competitive condition Noncompetitive condition
Experiment r 95% CI r 95% CI

CGC1 -.52 -.76, -.13 -.12 -.50, .30
CGC2 -.48 -.72, -.13 .13 -.25, .47
CGC3 -.29 -.56, .05 -.42 -.65, -.10

Confining this analysis to the PIF returnees grew out of the observation in the first experiment that this subset of participants demonstrated a stronger negative relationship between scoring and anxiety in competition. The selection of this subgroup was simply the consequence of our intention to examine several other variables that were obtained from information on the PIF. Our interpretation of the observation was that the PIF returners may have been a self-selected group of more motivated or interested individuals who reacted differently to the competitive situation. It is also possible that these participants were more compliant or had a greater desire to please the experimenters. In any case, it is instructive to compare these two groups using the larger number of participants in the combined data.

Figure 3 shows the Pearson correlations with 95 percent confidence intervals for PIF returners, nonreturners, and all participants. The combined results for PIF returners in the competitive condition yielded r(84) = -.41, p = .0001. In the noncompetitive condition, PIF returners also showed a negative correlation, but it was not significant, r(84) = -.12, p = .26. Nonreturners in the competitive condition yielded a negligible positive correlation, r(38) = .05, p = .75. In the noncompetitive condition, nonreturners produced a nonsignificant negative correlation, r(38) = -.13, p = .42. With both groups combined, the negative correlation in the competitive condition remains significantly negative, r(124) = -.25, p = .004; and the correlation in the noncompetitive condition remains largely unchanged, r(124) = -.12, p = .19.

One disadvantage of this competitive dice game that became apparent early in our research is that there is no demand on the participants to produce scores that are different from chance. The object of the game is to defeat the opponent, and this can be done by a single point. What is worse, the participant can achieve his or her goal quite dramatically without either score being above chance, and, most important, the participant had no idea of what chance expectancy was. Despite the absence of a clear context for significant departures from chance, it was apparent from the first two experiments that, at least among the PIF returners in the competitive condition, those who rated themselves as less anxious produced scores above chance whereas those who rated themselves more anxious produced below-chance scores. To examine this finding further we looked at the scores for the high- and low-anxiety participants in both conditions in the combined data of the three experiments.

The mean SAI for all subjects in both conditions was 18.07 (SD = 5.16). The SAI scores for the noncompetitive and the competitive conditions were not significantly different: noncompetitive SAI mean = 17.86, SD = 5.38; competitive SAI mean = 18.28, SD = 4.93; t(125) = -0.91, p = .36. The means for the various subgroups of interest ranged from 17.44 to 18.28. We therefore decided that the most consistent way of treating the data was to define high anxiety as SAI |is greater than or equal to~ 19 and low anxiety as SAI |is less than or equal to~ 17. Figure 4 shows the mean scores and 95 percent confidence intervals for high- and low-anxiety participants in the two conditions. These data are given for PIF returners and for all participants. Table 2 presents mean game scores, standard deviations, and single-mean t tests.

 PIF returners All participants

 Noncompetitive condition

High anxiety 200.36 204.65
 SD = 48.5 SD = 45.10
 t(30) = -0.97, p = .34 t(47) = -0.64, p = .53

Low anxiety 216.57 217.59
 SD = 46.85 SD = 49.00
 t(50) = 1.03, p = .31 t(72) = 1.35, p = .18

 Competitive condition

High anxiety 192.97 203.61
 SD = 35.82 SD = 41.48
 t(38) = -2.77, p = .009 t(58) = -0.97, p = .34

Low anxiety 212.82 220.36
 SD = 36.16 SD = 39.96
 t(38) = 2.07, p = .05 t(56) = 2.00, p = .05

Note: Mean chance expectancy is 209.34. All t tests are


The test used in this experiment was designed to mimic a real-life situation in which the anxiety of facing a competitive challenge might influence the manner in which subjects used PK to obtain the results in the true-RNG-based task. Two previous experiments demonstrated that high anxiety on entering the competitive game was associated with poorer performance (in terms of game scores) whereas low anxiety was associated with better performance. The experiment reported here confirmed that relationship. When the results of all three independent experiments are combined by the method of adding t's (Rosenthal, 1984), the result is z = 4.1, p = .00004, two-tailed.(7)

With some confidence that this relationship is not a chance effect, we can only reiterate a point made in our earlier paper, namely, that the manner in which psi effects are demonstrated by participants is likely to depend heavily on their individual perceptions of the test situation. The picture that has emerged from this series of experiments, best seen in Figure 4, is that the participants' anxiety on entering the competitive game seems to affect the direction in which the scores will go, irrespective of the overt instructions or stated goal of the task. Thus, the general effect of anxiety in this situation is not to "suppress" the deployment of psi ability, but rather to control its direction. It is not unreasonable to assume that this may reflect the operation of unconscious needs of the participants, but the specific connection must remain a source of speculation for the present.

That the trait anxiety scores so closely mirrored the state anxiety scores in this experiment at least suggests that a general trait factor may be contributing to these results. The trait test, however, had no precursor in the prior experiments, and so it remains for future research to determine the relative contributions of state and trait factors to the observed effects.

None of the other effects related to other variables proved to be sufficiently consistent across the three experiments to draw any general conclusions. As in the second experiment, we failed to find any relationship between competitive scores and the practice of a mental discipline. The negative correlation between SAI scores and game scores in the noncompetitive condition found in this experiment was not at all present in the second experiment and only weakly indicated in the first. Combining the experiments suggests a tendency toward a negative relationship between scores and anxiety even in the noncompetitive condition, but it is too weak and variable to be significant. One possible interpretation is that the competition enhances or accentuates a general effect of anxiety in test situations.

This experiment has been cast as one involving PK on a probabilistic electronic system. Since this series began, May, Radin, Hubbard, Humphrey, and Utts (1986) have advanced an alternative interpretation of RNG experiments that is based on a precognitive process. In the alternative interpretation, the subject uses precognition (or something like it) to "see ahead" in the stream of random numbers and initiate the trial at just the right moment to capture favorable subsets of the undisturbed stream of random numbers. This experiment, in which each trial of two digits was initiated at will by the participant, is well suited to a precognitive data selection interpretation, but the experiment itself cannot distinguish between the competing interpretations. Whether this experiment is ultimately best described as a PK disturbance effect or as a precognitive information process does not alter the fundamental interpretation of these results, which is that the effectiveness of that process will be subject to the participant's level of anxiety (and, therefore, perhaps other situational variables) in the test situation.

1 The development of the psi-test version of this game was supported by a grant from the Parapsychology Foundation, Inc.

2 The authors wish to thank Dr. Donald Burdick for his help in deriving these values.

3 The same situation existed for the first two experiments of this series, though in those cases we did not exclude the few participants who turned in the PIF late.

4 Unless stated otherwise, all probabilities are two-tailed.

5 Had we elected to continue with Spearman correlations, the result would have been: |r.sub.s~(33) = -.32. p = .03. one-tailed.

6 Specifically, the control data for PIF returners in the competitive condition produced |r.sub.s~(33) = .24, and in the noncompetitive condition, |r.sub.s~(33) = .05.

7 The relevant correlations are: CGC1, |r.sub.s~(21) = -.52; CGC2, |r.sub.s~(26) = -.49; CGC3, r(33) = -.29.


BROUGHTON, R. S., & PERLSTROM, J. R. (1986). PK experiments with a competitive computer game. Journal of Parapsychology, 50, 193-211.

MARTENS, R. (1977). Sport competition anxiety test. Champaign, IL: Human Kinetics Publishing.

MAY, E. C., RADIN, D. I., HUBBARD, G. S., HUMPHREY, B. S., & UTTS, J. M. (1986). Psi experiments with random number generators: An informational model. In D. H. Weiner & D. I. Radin (Eds.), Research in parapsychology 1985 (pp. 119-120). Metuchen, NJ: Scarecrow Press.

ROSENTHAL, R. (1984). Meta-analytic procedures for social research. Newbury Park, CA: Sage.

SPIELBERGER, C. D. (1973). State-trait anxiety inventory for children: A preliminary manual. Palo Alto, CA: Consulting Psychologists Press.

SPIELBERGER, C. D. (1983). Manual for the state-trait anxiety inventory: STAI (Form Y). Palo Alto, CA: Consulting Psychologists Press.
COPYRIGHT 1992 Parapsychology Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1992 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:psychokinesis
Author:Broughton, Richard S.; Perlstrom, James R.
Publication:The Journal of Parapsychology
Date:Dec 1, 1992
Previous Article:Honorton the meta-analyst.
Next Article:Contra George Hansen's flawed critique of the work with B.D.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters