CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
A "Psychic Contest" Using A Computer-RNG Task In A
Non-Laboratory Setting
Mario P. Varvoglis
Interface-Psi
Abstract
An exploratory, computer-controlled RNG study conducted in a
non-laboratory setting and represented as a "psychic
contest" is described. The study was undertaken to examine
whether the Psilab II "Volition" program could be profitably
used to explore intentional or nonintentional psi in the
setting of a "psychic fair". Sixty two subjects were
selected out of a larger population, on the basis of their
z-score in a preliminary psi test. Subjects were allowed up
to two Volition games each; 118 games were collected over
the three day period of the fair. An equal number of
"simulation" games, in which no subjects were present, was
also collected. In each game, both "feedback" RNG samples
(determining the progression of the feedback display), and
"silent" samples (which do not affect the game's feedback
display) were stored. Feedback and silent data were compared
to theoretical distributions through goodness-of-fit tests,
using end-game scores (z-scores) and within-game scores
(run-scores) as entries. The analysis of end-game scores
yielded no significant feedback or silent results, but the
silent run-score was significant (chi-square (28) = 47.03,
p =. 01). Neither the matched-simulation experiment, nor a
subsequent series of 100 extended-simulation experiments
showed any evidence for RNG malfunction.
The silent run-score result replicates the findings of two
prior Volition experiments by Berger (1988), and suggests
the feasibility of employing well-standardized computer
psi-tests for "field" investigations of psi.
36
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
Introduction
In the last decade, computerized psi tasks have become
increasingly popular in research laboratories, gradually
replacing the standard tools of prior generations, like
Zener cards, dice, and stand-alone random number generators
(RNGs) . This trend is largely due to the fact that computers
enable considerable experimental control, while introducing
previously unimaginable flexibility in hypothesis testing
and data analysis. Further, the trend toward computerized
psi-tasks reflects a growing interest in inter-laboratory
cooperation.
The release of "PsiLab II" (Berger & Honorton 1984;
Psychophysical Research Laboratories (PRL), 1985), a
standardized computer-RNG psi-testing system, has introduced
a new level of sophistication in collaboration and
replication efforts. One major advantage of PsiLab II is
that, because of its standardization (e.g., in hardware,
data-collection protocols, and subject feedback), it allows
for systematic comparisons of results across different
investigators and subject populations. Furthermore, because
of its portability and built-in safeguards, Psilab II can be
considered a self-contained "laboratory", i.e., a
transportable testing environment which can be taken outside
the laboratory to potentially promising environments.
The current study constitutes the first known attempt to
utilize Psilab's automated computer-RNG tasks under
circumstances quite removed from those of laboratory
research. The occasion was a 3-day conference in Montreal,
where I had been invited to give talks on psi research. In
addition to the formal presentations, there was a "psychic
fair", with holistic health merchants, New Age artists,
tarot-readers, palm-readers, astrologers, past-life
regressors, and other colorful personalities. It seemed to
be an interesting setting for a psi experiment, and, about a
month prior to my arrival, I proposed creating a "psychic
contest" for the fair. The organizers were overjoyed with
the idea (thinking, no doubt, of the associated publicity)
and agreed to rent out a sizable booth at a discount.
The "contest" involved two tasks, each involving a separate
computer. The first, a computer psi-game I created for the
occasion, served as a screening/motivational device. The
second, the "official" psi task, was Psilab's "Volition"
game. Volition is a computer psi game experiment in which
subject-initiated button presses sample the RNG. Each button
press (run) samples 100-bits of RNG data which drive a
graphic feedback display. Another 100-bits, designated as
"hidden" or "silent" data, are also sampled but not
displayed to the subject who is blind to these data. I chose
Volition partly because, from among several available
37
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
choices, it seemed the easiest to explain, in a,hurry, to a
subject "off the street"; "and partly because it has already
been used in a number of studies, with some success.
Prior research with Volition
The first Volition study, conducted at Psychophysical
Research Laboratories [PRL] (PRL, 1984) involved 20
participants, each contributing 10 "games" (with 100 runs
each containing 100 bits). Overall, there was no evidence
for psi in the "feedback" samples, but a significant excess
of subjects obtained independently significant results in
the "silent" data. The silent effects were non-directional :
game outcomes deviated significantly from chance, but not
consistently with the person's "aim".
Palmer & Perlstrom (1987) reported a Volition study with 30
subjects, examining the effect of different instructional
sets (instructions emphasizing directional control vs.
extremeness of scoring) . Results from this study are
difficult to interpret, due to the multiplicity of analyses
undertaken, but the most salient finding seemed consistent
with PRL results: game-score variance in the silent samples
was relatively high with instructional sets for "extreme"
scoring, and relatively low when subjects were aiming for
the "chance" line.
Two more Volition studies, reported in Berger (1988), and
based upon run-score variability (rather than game-score
variance) produced conceptually similar results. In the
first, involving 10 subjects (including the investigator)
significant run score variability was obtained in the silent
data, whereas no effect was evident in the feedback data;
removal of the experimenter's data did not substantially
change the results. In the second study, in which the
investigator was the only subject, significant run-score
variability was again found in the silent, but not the
feedback samples. These results with Volition replicated the
results of two earlier studies by Berger (1988), using
similar RNG sampling procedures but based on a different
feedback task (PsiLab's "Psi Invaders) .
In short, Volition has consistently shown some promise for
eliciting non-directional silent data effects. Naturally,
given that many RNG-feedback studies have demonstrated
intentional psi, one wonders whether there is something
special about Volition (and other tasks which include a
"silent" condition) which specifically invites unintentional
psi effects. Does the mere existence of a silent condition
distract from the intentional task and invite
"displacements"? Schechter (1987) reported data supportive
of a "displacement" interpretation: individuals who tended
to "miss" in the feedback task (i.e., to obtain end-results
contrary to their chosen aim) tended to "hit" in the silent
task.
38
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
On the other hand, it is also possible that silent effects
are, in fact, no more than experimenter effects. First, the
investigators' own psi could be shaping the silent data -
during the session (through psi-mediated data sorting or
through "conformance behavior"), or retroactively (as
suggested by Observational Theories). Alternatively, the
investigators' expectations may create tacit "demand
characteristics" in the study, which unconsciously influence
subjects' psi performance. The reported Volition studies
have been based upon intensive laboratory work with
self-selected volunteers -- people who have prior interest in
psi (and in psi research), and who are given a fair amount
of attention prior to, and during the testing (through
repeated laboratory visits, interchanges with lab members,
extended task-explanations and demonstrations, etc. ). Under
such circumstances, it is plausible to believe that subjects
might simply "give the experimenter what he wants" - if not
feedback effects, at least silent effects.
The question of the experimenter's role is particularly
pertinent when it comes to automated tasks like Volition.
Such tasks hold promise as self-contained, experimenter-
independent procedures. But to be used in this manner, they
must be motivationally (and not just methodologically) self-
contained; they cannot depend too much upon inspirational
investigators and special interpersonal settings.
The "psychic fair" Volition contest thus seemed to be a way
to determine whether effects similar to those already
reported would be obtained in situations in which
individuals' motives for participation and interactions with
the investigator are quite different from those typical of
laboratory research. Though participants would still
interact with the investigator, and could not be considered
100% "off the street" (not in a psychic fair!), still,
several factors rendered the setting much closer to the
"real world" than to the world of the laboratory. To mention
a few: the billing of the psi test as a contest, the
market-place ambience of the "psychic fair", the necessarily
brief (and business-like) subject-experimenter interactions,
the concrete possibility of winning a prize, and, above all,
the stiff price each person had to pay to have a shot at it!
Subjects
Because it was impossible to know, in advance, how many
individuals would be drawn to the "contest", and how many
would meet the screening criteria set, the number of
subjects could not be defined in advance (though an upper
limit of 100 subjects was set). To avoid accusations of
"optional stopping", the limits of the experiment were
defined temporally: it was decided to run all subjects
meeting the screening criteria, from the opening of the fair
39
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
until closing time each day (i.e., 10:00 PM) .. Each subject
would be allowed a maximum of two Volition games.
Generally, subjects either came purposefully to the testing
booth, after having read notices for the contest elsewhere
in the fair, or wandered in, attracted by the crowds and/or
the computer displays. In the three days of the fair, over
220 people paid to take part in the screening task. Of
these, 62 participants (information on gender breakdown was
not retained) met screening criteria, and were willing to
pay the extra fee to participate in Volition. With the
exception of one individual, who was a fellow psi
researcher, none of the participants had been formally
tested for psi (until then); of course, many of them may
have had spontaneous experiences or tested themselves
informally, but this was not explored.
Setting
The experiment took place in one of the booths set up for
the "Sommet Esoterique' at the Velodrome Olympique of
Montreal. Because the environment was quite bright and
noisy, a special tent was constructed with dark fabrics,
closing in the testing area on three sides and on top. The
area under the, "tent" was about 12 meters square.
The two computers used were placed at right angles to each
other, on separate tables, with the color computer facing
the opening of the tent.
Hardware
An Amiga 1000 with a color screen, a "mouse", two disk
drives, and a 2 megabyte memory extension was used for the
preliminary screening task. For the Amiga, no hardware RNG
was used; the random digits were based upon an algorithm,
reseeded by the computer's clock.
An Apple Ile with a green/black screen, two disk drives, two
"paddles" and a printer was used for the official psi task.
The source of random digits for the Apple psi task was a
Psilab II noise-based RNG, fitted into Slot 5 of the Apple.
This RNG had been given to the author in 1985 by PRL staff,
after having passed a battery of tests ensuring its proper
operation. A detailed presentation of the PRL component
integrity tests, safeguards (such as shielding), and
randomness checks is available elsewhere (Berger & Honorton,
1984; PRL, 1985).
40
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
-

SoApproved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
Volition: In Psilab's Volition test, subjects are provided
with continuous visual feedback as to their cumulative RNG
scores through a graphic line which moves across the screen,
in short segments. Each time the subject presses the button
of the Apple game paddle, a 100-bit RNG sample is taken and
compared to an alternating "target" bit stream; this yields
a run-score with a mean chance expectation of 50 and
standard deviation of 5. The run-score is added to previous
scores and the cumulative z-score calculated; this
determines the direction (upward or downward) and slope of
the new feedback-line segment. Thus, above chance scores
tend to direct the feedback segment upwards, below chance
downwards. With the help of trend lines demarcating chance,
and plus and minus 2- and 3- standard-deviation thresholds,
the evolving feedback line represents clearly the cumulative
performance of the person at each moment.
In parallel to the feedback RNG runs, each buttonpress
results in a 100-sample silent run, as well. The designation
of relative order (whether the first of the two samples is
"feedback" or "silent") is alternated on a run by run basis.
The Volition task used in the study was practically
identical to that described in full in Berger & Honorton
(1984), and Berger (1988). Only two differences were
introduced. First, through the Design option, the game
length was set at 20 100-sample RNG runs (in contrast to
other investigators' setting of 50 or 100 runs). Second, at
the beginning of each game, subjects were only asked if they
prefer "Hi-aim" or "Lo-aim". They were not offered any other
options for "tailoring" the feedback to their preferences;
these options had been set previously (with "graphic
designs" off, and all other options on).
Buddha Game: The Buddha Game was written for the Amiga
computer, in the C language, by a programmer who followed
the author's instructions. As in Volition, the subject's
buttonpress results in a series of random bits; the subject
attempts to "sense" the right moment, so as to obtain the
maximum run-score possible. Unlike Volition, however, the
random bits are not obtained from a hardware RNG, but are
derived from the built in Amiga random function, "reseeded"
by a digit from the Amiga clock.
Essentially, the game consists of a series of digitized
images depicting a golden Buddha statue surrounded by an
electric blue aura. Depending on the random score obtained,
the buddha image either grows in size (giving the impression
of an advance toward the user) and then turns clockwise, or
turns counter-clockwise and then diminishes in size (giving
the impression of a retreat). Accompanying these movements
is a digitally sampled sound, vaguely resembling "Ahhhh",
which decreases in pitch with "advances" and increases in
pitch with "retreats".
41
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
At the beginning of the game, the Buddha is positioned at
the "mid-point", half way from the first and last images.
Once the subject presses the Amiga's left "mouse" button,
the RND function is sampled 10 times, yielding a series of
1's and O's. If the runscore is over 5, the Buddha advances,
if under 5, he retreats, and if at 5 he stays stationary.
The greater the departure from the expected score, the
greater the advance or retreat from the current position.
The goal of the individual is to make the Buddha either
advance or retreat consistently, so that he reaches either
of the two end points. The complete game consists of 22
buttonpresses (runs) . Following the last run, a sampled
(digitized) sound of children laughing is heard, the screen
goes blank, and the overal game z-score is displayed.
At the time of the fair, this program was not finished. No
procedure for entering subject names, or for storing subject
data had been implemented, and there was no provision for
control runs. Thus, I decided in advance that this game's
outcomes could not be used to assess psi performance;
instead, they would just serve as a motivational "prop" for
Volition, i.e., as a means for persuading the person that
they are ready for the "real" test.
Procedure
As it turned out, the contest was the most popular event of
the fair, and our booth was literally deluged with people
crowding around, waiting for their chance to test their
psychic muscle. The unanticipated popularity of the contest
resulted in a a rather hectic atmosphere, clearly removed
:
from the sanguine, well-disciplined atmosphere of the
laboratory. Though an effort was made to keep the situation
under control, some variations in testing conditions and
experimenter-subject interactions were inevitable.
Upon arriving at the tent, people would either read the
posted explanations of the contest, or would inquire further
as to what's going on. If I was momentarily available, I
would briefly explain the general idea; otherwise, I would
direct individuals to the posted explanations, and ask them
to await their turn, for more details. (No attempts were
made to solicit participants; it was completely unnecessary,
at any rate). In general, Volition was presented as the
"official" task to which subjects had to "graduate": they
first had to participate in the Buddha game, and obtain a
minimum z-score of 1, in order to qualify for Volition.
If interested, the person would pay the cashier the fee for
the Buddha game ($3.00), and a ticket would be given, with
the word "Buddha" and the corresponding fee written on it,
as well as the person's name, address and phone number. The
receipts were numbered, and as soon as one participant
42
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
finished with the Buddha game, the next one would be called
by number, and sit in front of the Amiga screen. I would
then explain the Buddha game. There were some variations in
instructional set, from subject to subject, as some people
had already been there for a while, and had seen several
demonstrations, while others were newcomers. Generally,
subjects were told that the Buddha game tests their
intuition, their ability to "sense" the right time for
pressing the button, in order to obtain high scores. I used
the analogy of a fast-spinning roulette wheel, with numbers
on it, which the subject stops, through his button press; if
they stopped it, say, on "odd" numbers, then the Buddha
would advance, if on "even", he would retreat. It was
stressed that the goal is to be consistent in finding "odd"
or "even" numbers, and that the degree of consistency would
be signified by the Buddha's progress in one particular
direction (advancing or retreating). I then showed the
subjects how to use the "mouse", and stayed next to them for
the first few trials, until I felt they understood the
relationship between the Buddha movements, and their scores.
Following these instructions, I would either move back, and
join the crowd behind the Buddha game player, or would turn
to the next Volition player, i.e., the person who had
already passed the Buddha game, and was waiting for me to
start Volition. Meanwhile, the Buddha game player would go
through the psi task alone, pressing the mouse-button
repeatedly until the game ended, and the final z-score was
displayed. I marked the score on their receipt, and then
gave the person some feedback, modulating my comments
according to the absolute z-score. If the score was below 1,
I would generally reassure subjects that they were
undoubtedly much more intuitive than the score suggests; but
then I would add that the contest procedure demands a
minimum score of 1 to continue. (In a few cases, in which
the z-score was over .9, and in which I sensed the person
was greatly disappointed that they had "just missed the
mark", I made an exception and allowed them to enter the
Volition test). With absolute z-scores of 1 or higher, I
generally created quite a fuss (the higher the score, the
greater the fuss), and concluded by telling subjects that
they could now participate in the contest, if they wanted
to, but that they were not in any way obligated to do so.
If they did decide to continue with Volition, they went to
the cashier, who collected the appropriate fee ($4.00), and
marked the word "Volition" on the receipt. The participant
would then wait in the Volition queue, or come directly to
me, if no one was currently playing Volition. At this point,
I would start the Volition session by typing in my
three-character password (these characters are not displayed
on the screen), and then register the participant by name.
Under "participant ID", I would type in the absolute z-score
obtained in the Buddha game; this, however, was only done
43
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
after the first few sessions had been completed (thus Buddha
game scores are missing for 7 subjects) .
I introduced Volition by stating that this test was somewhat
more challenging than the Buddha game, but that their score
on the Buddha test showed they were "up for it". I added
that Volition was also more accurate: the person could trace
his scoring patterns with great precision, and use these to
test mental strategies. It was also explained that, whereas
the Buddha game was strictly based on intuition, here, one
could alternatively use a "mental force" (i.e., PK) approach
and "oblige" the line to move in the desired direction.
When the Volition "aim" question came up, I used the analogy
of "heads" or "tails", in a coin toss, to convey that
subjects could choose either "hi-aim" or "lo-aim". However,
I also stressed that this was merely a focusing device; if
the feedback line insisted on moving in the direction
opposite to their choice, they should just "go with it", and
try pushing it even further in that direction. I emphasized
that the winner of the contest would be the one whose
feedback line departed maximally from the baseline,
irrespective of aim.
I then would input subjects' "aim" choice myself, using the
game-paddle, and would hand them the paddle when the
complete Volition display had been drawn on the screen. The
"mode" for all subjects' first game was "manual". Subjects
were urged to press the button once, so they could see the
first segment of the feedback line, and understand its
movement in relation to their "aim" and the baseline. Then I
would leave them on their own. Following completion of the
first game, I commented on the score; again the higher the
absolute z-score, the greater the compliments. With low
absolute z-scores (below 1) I sought to point to something
promising in the feedback curve and attributed declines to a
loss of concentration.
In all cases, I offered subjects a second opportunity,
stating that they were entitled to a maximum of two games,
with the best score of the two being used for the contest.
The great majority of participants did indeed choose to play
a second game. All were again asked for "high" or "low" aim;
then the subject proceeded, as in the first game, using the
"manual" mode. (In the case of 2 subjects, after having
observed their frustration in the first game, I suggested
they try the "automatic" Volition mode, to see if their
scoring would improve). In cases where subjects had high
absolute z-scores (over 1.8) in either of the two games,
they were told to make sure they return for the closing
night of the fair, when the winners would be announced.
Toward the end of each day (around 10:00 PM) the cashier was
instructed to stop accepting payments for the Buddha game.
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
After "running" the remaining subjects, the equipment was
turned off, and the front of the tent closed. The equipment
was left in the tent overnight, but I took the Voltion
program and data disks home with me. Guards were present in
the area of the booths the entire night (as all vendors
would leave their merchandise there), and one of the
organizers slept in our tent, to ensure the safety of her
Amiga (which she had lent me, for the screening test).
At the end of the third day, all z-scores and subject names
were printed out on a sheet, and the highest absolute
z-scores singled out. With the help of the organizers of the
fair, we announced the winners of the contest, and invited
them to come collect their prizes. In instances where a
winner was not present, the individual with the next highest
z-score was called. This continued until the first prize (a
small Canon computer) and three second prizes (some posters)
had been distributed.
A lapse in protocol occurred in one game, and I was forced
to be the subject because I accidentally started the game
myself. As mentioned earlier, I would set subjects' aim.
High aim is selected by turning the paddle knob fully
clockwise, and then pressing the paddle button. However,
this knob setting also sets the game which follows on
"automatic" mode, whereby the feedback line immediately
.
starts moving across the screen without any further button
presses. It was because of this that I always input the
subjects' "aim" preferences (hi-aim, in the vast majority of
cases) myself. However, in this one instance, I must have
been somewhat fatigued, because I forgot to immediately turn
back the paddle knob, counter-clockwise, just after inputing
the subject's aim. The Volition display came on, and, as I
was preparing to hand over the paddle, I saw (dumbfounded)
the feedback line move all by itself. I immediately turned
the knob counter-clockwise, but the damage had already been
done, and a few runs had definitely accumulated, moving the
feedback line in the wrong direction. Passing this situation
over to the subject (who was dreaming about the first prize)
would have been in poor taste, so I was forced to complete
that game myself. It turned out to yield the highest
absolute z-score in the experiment (-2.68) .
Simulations
Simulation games: The second night after the closing of the
psychic fair, I initiated a series of matched "simulation"
games, provided with all Psilab Il software. Generally,
these simulate sampling and timing conditions of the game,
but without a player pressing the button, and with no image
on the screen. The Apple and RNG were situated in the room
in which I was staying, and the simulation, involving a
total of 118 games (the number of contest games accumulated
45
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
over the 3 day period) took place overnight, while I was
sleeping. Due to the logistical constraints, it would have
been impossible to run the simulations in situ.
Extended Simulations: Psilab Il comes with two Random
Analysis programs - the Frequency Analyzer and the Serial
Analyzer. Prior to undertaking the present study, it was
attempted to run both of these programs; neither of them
worked. Later, however, an alternative approach was
suggested by Berger, who kindly provided the software
necessary to perform a series of "extended simulations"
which could serve as an empirical background against which
the present experiment and matched- simulation could be
juxtaposed. Quoting from Berger (1988) :
"Extended simulations are composed of the equivalent amount
of data as a complete experiment (as contrasted with the
matched game simulations which each have the equivalent of
one game's data) ... For the extended simulations, the RNG
sampling software was extracted from [the game program] and
the quasi-random inter-trial latencies produced by subjects
in experimental games was replaced by a fixed-speed sampling
regimen operating at the full speed capable of the Applesoft
BASIC language. Extended simulations test both the integrity
of the hardware and software used in the experiments, as any
systematic biases in either should be magnified" (in press) .
Hypotheses and planned analyses
Previous Volition research unequivocally pointed to
non-directional silent-condition effects (either at the
run-score or terminal z-score level). Implicitly, the
experimenter's expectation was that similar effects might
turn up in the present study. However, no explicit
predictions were made. Because of the unusual
data-collection circumstances, it seemed more appropriate to
adopt a "wait and see" attitude; the study seemed best
conceived as exploratory, rather than as a replication.
Nevertheless, certain specific analyses were planned.
Insofar as both end-game and run-score measures have shown
promise in past Volition studies, both were used as
dependent variables. Each measure was assessed through a
goodness-of-fit test.
Run-score variability was examined through a chi-square
goodness-of-fit test, identical to that utilized by Berger
(1988) in his own Volition studies. Essentially, this test
involves comparisons of the observed frequency of each
run-score value (e.g., 48, 49, 50, 51, etc.) with the
expected frequency for that value.
1. Failure was due to an incompatibilitywith the printer hardware, which
has since been corrected
46
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

.
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
Cumulative (terminal) z-scores were examined through the
Kolmogorov-Smirnov [KS] goodness-of-fit test, provided by
the PRL analysis-software and presented in the Psilab Manual
(1985), and more fully in Knuth (1981, pp.45-58). The choice
of the KS was prompted by the suggestion, in the Psilab
Manual, that in assessing large amounts of randomness data
the KS may be preferable to other, more commonly used
statistics. It seemed that, insofar as the KS is sensitive
to both local and global departures from theoretical
expectation, it could be simultaneously used to examine the
adequacy of the RNG, and the presence of any consistent
scoring patterns, on the part of the subjects.
Essentially, the KS compares the distribution of the
observed z-scores against their expected distribution. The
degree of "fit" between the empirical and theoretical
distributions is summarized by two statistics, K+ and K-,
representing the average deviations of the empirical curve
below and above (respectively) the theoretical distribution.
Results
Table 1 summarizes the results at the game-score level,
based upon 118 z-scores for each of the four conditions.
Depicted are the mean z-score, and the K+ and K- statistics
of the KS goodness-of-fit tests. As can be seen from the
p-values of Table 1 no significant departures from
theoretical z-score distributions were obtained for feedback
or silent data, in either experimental or control
conditions.
Following Schechter (1987), each Volition game was
classified as a "miss" or a "hit" according to the feedback
z-score. Using appropriate t-tests, the mean silent z-scores
for each type of game were compared to chance and to each
other. Both mean silent z-scores were at chance ( "miss"
silent mean z =. 163, t(61)=1.195, ns .; "hit" silent mean
z =. 042, t(54) =. 343, ns). The difference between "hit" and
"miss" silent data was not significant (t (115) =. 655, ns).
The run-score results are graphically represented in Figure
1 (1a and 1b for experimental data, 1c and 1d for matched-
simulation data); the frequency of each runscore is plotted
against the theoretical baseline (the z=0 line). Table 2
Summarizes results from the chi-square goodness-of-fit test,
based upon 2360 runs (118 games x 20 runs) for each
condition, and comparing the distribution of run-score
values for all cells between 36-64 (inclusive) to the
binomial theoretical distribution. (Given the number of
observations involved, the expected frequency below 36 and
above 64 was too low for a chi-square analysis; tail-end
cells were collapsed, to maintain expected frequency above
5). As may be seen from Table 2, the goodness-of-fit
analysis shows significantly high variance for the silent-
experimental conditions (chi-square [28 df] = 47.03, p =
.01). This translates to an effect size of .045 (by
47
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

-
48
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
converting the probability value to a one-tailed z-score,
then dividing the z-score by the square root of n). This is
more than double the magnitude of the effect sizes of the
Berger (1988) studies (calculated to be .008 and .02).
As shown in Figure 2, the result of 100 extended simulations
for each condition ( "feedback" and "silent") showed no
excess of significant chi-square results. When the
experimental and matched simulation data are juxtaposed
against the extended simulations, we see that, while the
matched-simulations showed good overall randomness, the
experimental silent data lay in the tail-end of the
distribution.
Discussion
The purpose of this exploratory study was to use a
well-standardized computer-RNG task to examine subjects' psi
performance in a setting quite different from that of the
laboratory. The idea was to determine whether or not results
would be consistent with those found in prior research, in
view of large differences in subject incentives,
subject-experimenter interactions, and general ambience
during testing.
As suggested by the KS analyses, there were no significant
departures from the expected distribution of terminal
z-scores. Despite the (presumably) strong incentive value of
a high end-game score, subjects were apparently unable to
"push" the feedback line to a final result consistent with
their goals (i.e., a large z-score). In this respect, the
null end-game feedback results are similar to those reported
in all previous Volition studies. On the other hand, insofar
as there was no evidence for a silent effect at the level of
the cumulative z-score, the present study does not replicate
the PRL findings (showing some evidence for bidirectional
scoring in the silent data). Nor do the results provide any
strong support for the idea that feedback "missers" were
silent "hitters"; though the trend was clearly consistent
with that reported by Schechter (1987), it did not in any
way approach significance. (It should be noted, however,
that the present study's instructional set, emphasizing
"extremeness" of scoring, was quite different from that of
the PRL study, emphasizing directional scoring. )
At this point, it seems safe to state that, in tasks such as
Volition, the researcher should not focus exclusively upon
the end-game score to assess psi performance; at the very
least, run-score measures should be included. As Berger
(1988) has argued, in tests allowing for multiple
subject-interventions, the most immediate "unit of effort"
is the "buttonpress", i.e., the run. Many subjects who may
not be able to maintain consistent performance, may
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
nevertheless show short-lived performance "peaks",
detectable at the level of the run-score.
This, at least, is suggested by the results of the present
study. As shown in Table 2, while the matched-simulation
run-score data showed good fit to the theoretical distribu-
tion, the experimental silent data were significantly
deviated from chance. Given that the extended simulations
also showed the adequacy of the RNG's operation, it seems
safe to state that the observed silent effect was probably
due to psi, and not to some software or hardware artifact.
The silent result thus replicates the findings of Berger
(1988), who obtained similar run-score effects in the silent
data of two Volition studies, as well as in two other
studies (using Psi Invaders, another Psilab program).
Indeed, the effect size of the silent result of the present
study was considerably larger than those of the two Berger
studies. Perhaps the psychic contest situation somehow
created a psi-conducive dynamic (which, unfortunately, did
not manifest in the explicit task!). Also, it is possible
that the screening procedure - the Buddha game - heightened
the expectations of those who made it through into Volition,
and thus contributed to silent scoring.
In general, the present Volition results are conceptually
consistent with those of a number of studies, showing more
pronounced effects in silent or non-feedback RNG data than
in feedback data (Berger, Schechter & Honorton, 1986; Braud,
1978; Palmer & Perlstrom, 1987; Varvoglis & Mccarthy, 1986) .
Insofar as the present experiment took place in a social -
psychological context quite removed from laboratory
settings, the results lend further support to the idea that
silent effects indeed reflect subjects' experience of the
task, rather than deriving from the tacit "demand
characteristics" in laboratory settings.
Nevertheless, much remains to be done to adequately
demonstrate the independence of silent effects from
psi-mediated experimenter effects. Despite the unusual
testing circumstances of the present study, it clearly
cannot be considered a "stand-alone" experiment: there were
at least two major ways in which investigator-psi may have
shaped the results.
First, I myself may have contributed to the results during
the unfoldment of the experiment. Of course, I was observing
the progression of each game, and hoping for good outcomes.
Simultaneously, there were a (highly variable) number of
other observers, who, undoubtedly, were harboring mixed
feelings toward high scorers (i.e., potential competitors
for the first prize). Perhaps, at an unconscious level, I
suspected that the only way to get a decent result out of
this experiment was through the silent condition - while all
the competing observers were busy focusing on the subject's
49
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
feedback line. \The possibility of unintentional
experimenter-psi is certainly consistent with the fact that
I accidentally obtained the highest score in the experiment.
Second, if we accept the premises of Observational theories,
then we have yet someone else to blame for the silent
effects: Rick Berger. Following completion of this study,
and prior to any analysis or observation of the silent
results, I sent duplicates of my data to Berger, who had
kindly offered to do the run-score goodness-of-fit analyses
for me, using the programs with which he had analyzed his
Volition and Psi Invaders studies. Thus, in effect, Berger
was the first observer of the present study's silent data.
If we take the idea of retroactive-PK seriously, then it is
possible that the pattern obtained in the silent data is due
to Berger's psi, and not to the contest participants. In
such a case, obviously, the current study could not be
considered an independent replication of Berger's data -
just a further confirmation of his psi!
In any event, insofar as this is the fifth Volition study
showing some kind of silent effect, it encourages further
exploration of such automated psi tests. The next step, it
would seem, would be to collect psi data using a truly
"self- standing" system (complete with instructional set,
motivational devices, and no experimenter) while assessing
any "observational" experimenter effects through split-data
analyses. Perhaps such an approach would help us determine
whether apparently systematic "errors" in psi - displace-
ments, silent effects, field effects, etc .- are indeed
intrinsic to the motivational/informational characteristics
of the psi task (Varvoglis & Mccarthy 1986), or whether they
simply 'reflect investigators' and subjects' tacit
construction of the meaning of the experiment (Weiner, 1987) .
Table 1: Mean-Z scores and KS Summary statistics
i
Experimental
Simulation
1
Feedback
Silent
Feedback
Silent
mean-Z
-. 016
.092
. 108
-. 014
P
.648
.323
.233
.632
K+
.57
.22
.21
.70
P
.517
.906
.917
.378
K-
.37
.89
.89
.78
P
.762
.207
.205
.295
Table 2: Run-score distributions (2360 runs)
Experimental
Simulation
Feedback
Silent
Feedback
|Silent
chi-sq(28 df)
p
26.47
.55
47.03
.01
24.08
.43
28.57
.68
50
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
chi-sq[28 df]

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
Fig.1: Feedback and Silent run-score distributions
for Experimental and Matched-simulation data
FEEDBACK
SILENT
3
·
2
1
:
Z
Z
0
-1
-2-
-2
-31
.
33334444444444
455
1565655566666
67890123456789012345676901234
Run Score
Fig.la
3333444
050560065066666
67890123456789012345678901234
Run Score
Fig.1b
SIM FEEDBACK
SIM SLENT
3-
3
2
1
1
Z
2
0
-1.
-1
-2
-2.
-3J
-31
3333444444
6856055666666
67890123466789012340676901234
Run Score
Fig.1c
67890123466789012345678801234
Run Score
Fig.1d
1
Fig 2: Extended "Feedback" & "Silent" Simulations and
corresponding Experimental & Matched-simulation results
Extended "Feedback" Simulations
Extended "Silent" Simulations
chỉ-sq[ 28 df ]
chi-sq[28 df]
40
10
51
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3

Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3
REFERENCES
Berger, R. E. (1988). Psi effects without real-time
feedback. Journal of Parapsychology (in press).
Berger, R. E. and Honorton, C. (1984). PsiLab II: A
standardized psi-testing system. Proceedings of the 27th
Annual Meeting of the Parapsychological Association,
355-377.
Berger, R. E., Schechter, E. I. and Honorton, C. (1986). A
Preliminary review of performance across three computer
psi games. In D. Weiner and D. Radin (Eds. ) Research In
Parapsychology 1985 (pp. 1-3). Metuchen, NJ: Scarecrow
Press.
Braud, W. G. (1978). Recent investigations of microdynamic
psychokinesis, with special emphasis on the roles of
feedback, effort, and awareness. European Journal of
Parapsychology, 2, 137-162.
Knuth, D. E. (1981). The Art of Computer Programming: Vol.
2/Seminumerical Algorithms. Addison Wesley: Reading Mass.
Palmer, J. and Perlstrom, J. (1987). Random Event Generator
PK in relation to task instructions. In D.H. Weiner and
R. D. Nelson (eds. ) Research In Parapsychology 1986
(pp.17-20). Metuchen, NJ: Scarecrow Press
Psychophysical Research Laboratories (1984) . 1983 Annual
Report. Princeton, N.J .: Psychophysical Research
Laboratories.
Psychophysical Research Laboratories (1985) . PsiLab II
User's Manual. Princeton, N.J .: Psychophysical Research
Laboratories.
Schechter, E. I. (1987). Missing and displacement in two RNG
computer games. In D.H.Weiner and R. D. Nelson (eds. )
Research In Parapsychology 1986_ (pp.73-77). Metuchen, NJ:
Scarecrow Press
Varvoglis, M. P. and Mccarthy, D. J. (1986).
Conscious-purposive focus and PK: RNG activity in
relation to awareness, task-orientation and feedback.
Journal of the American Society for Psychical Research,
80, 1-30.
Weiner, D. H. (1987). Thoughts on the role of meaning in psi
research. In D.H. Weiner and R.D. Nelson (eds. ) Research In
Parapsychology 1986 (pp. 203-223). Metuchen, NJ: Scarecrow
Press
52
Approved For Release 2000/08/11 : CIA-RDP96-00792R000700610001-3