Feedback Considerations in Anomalous Cognition Experiments

APPLOVER FOR RSISASH ROIUNSINE SC IA-RDP96-00789RGR320022AMOM+a04

Feedback Considerations
in |
Anomalous Cognition Experiments

by

Edwin C. May
Nevin D. Lantz
Science Applications International Corporation
Cognitive Sciences Laboratory
Menlo Park, CA
and
Tom Piantineda
SRI International
Menlo Park, CA

Abstract

In order to determine from what time frame the data from anomalous cognition (AC) originates, we
have examined the role of precognition and feedback on the quality of AC. In an otherwise standard
AC protocol, we displayed feedback tachistoscopically to receivers. The cognitive awareness of the
feedback content was minimal, and 2 of the 8 intensities used for visual display of the feedback were
below the visual recognition threshold. We hypothesized a number of possible relationships between
feedback intensity and AC quality, including one based on precognition (i.e., the data originated from
the future feedback). Four viewers contributed 40 trials each (5 at each of 8 different intensity bands).
Using a sum-of-ranks statistic, two viewers produced independently significant evidence of remote
viewing (ie., binomial probability of 2 hits in 4 trails with an event probability of 0.05 is 0.014 ). None of
the data showed a significant positive correlation of feedback intensity with AC quality; one receiver
showed a significant negative correlation. This result is discussed with regard to precognition in general
and the troublesome unfalsifiability aspect of truly goal-oriented precognition.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 1

Lae ee ee reN ease MO eo ts CIA-RDP96-00789RODSPROPININ 1964

xperimentS

Introduction

Beginning in 1986, we conducted a 2-year investigation of the dependency of the quality of anomalous
cognition (AC) on the feedback to the participant; hereafter called the receiver.” The experiment was
conceptually quite simple, but to address precognitive issues the experiment became technologically
complex. Related to the general feedback issue, we were interested in determining from what time
frame a receiver gains access to the target.

One model of AC, which is based on precognition, is that the data originate from the future feedback to
the receiver (Feinberg, 1975). If we are to understand the process of AC from this perspective, we must
examine and understand precognition. One view is that precognition is simply a “clean” methodology
for conducting free-response experiments. For the purposes of this paper we have assumed that pre-
cognition is real. We cite Honorton and Farrari’s (1989) meta-analysis of the forced-choice experi-
ments from 1935 to 1987 as, at least, prima facie evidence for the phenomenon. Using 309 studies, they
found overall significance (Z = 11.4, p <2.4 x 10-30 ES = 0.020+0.002), and they examined a variety
of variables, including file-drawer and quality considerations.

Targ and Targ (1986) conducted an experiment to explore the relationship between feedback and pre-
cognition; do individuals have access to actualized or probable futures? Their evidence suggests that
precognition is independent of a priori target probabilities. However, in a similar experiment, Radin
(1988) subsequently found significant evidence contradicting their result.

In a carefully constructed experiment, Vassy (1986) found that a goal-oriented model of precognition
did not fit his data; however, our analysis (May, Utts, Spottiswoode, and James, 1994) demonstrated
that the timing parameters that Vassy used in his experiment could not exclude a goal-oriented inter-
pretation of the results.

We will discuss the apparent contradictory results of these experiments in the Discussion Section of this
paper. We note that neither the Targ and Targ nor the Vassy studies explored the functional relation-
ship between feedback and AC. Nonetheless, because of their more recent investigations and because
of the importance of the question about feedback, we are presenting an updated and more complete
version of our 1986 experiment than appeared earlier (May, 1993).

Conceptual Approach

In a schematic protocol for a typical AC trial, a receiver and monitor (i.e., interviewer) are sequestered
ina laboratory. An assistant randomly selects a target from a pre-defined set (i.e., target pool), andata
predetermined time, the receiver attempts to describe the target. At the end of the session, the AC data
are secured, and the intended target is shown to the receiver as feedback. Normally the feedback in-
cludes a presentation of the target and involves a complete debriefing of the session experience.

In our experiment, we attempted to inhibit as many potentially open precognition channels as possible
so that we could assess the impact of the visual intensity of the feedback display on the quality of the AC.
Thus, we eliminated all debriefing of the AC experience during the feedback periods and presented

* The Cognitive Sciences Laboratory has adopted the term anomalous mental phenomena instead of the more widely known psi.
Likewise, we use the terms anomalous cognition and anomalous perturbation for ESP and PK, respectively. We have done so
because we believe that these terms are more naturally descriptive of the observables and are neutral in that they do not imply
mechanisms. These new terms will be used throughout this paper.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 2

APRLOVER A QR SHSASR 2090/08/08,i.CIA-RDP96-00789RAGI2002ZNG08 1594

visual feedback tachistoscopically. The display intensities for the feedback presentation varied from
zero to a level that just exceeded the visual recognition threshold (i.e., some details can be recognized).
Even the strongest display intensity was insufficient to provide a “satisfying” study of the target material.

To lessen the impact of other potential precognitive leakage paths, extreme care was taken in order to
insure that the receiver was the only individual who was simultaneously aware of both the intended tar-
get and the response. That is, during the course of the experiment and to date, no individual except the
receiver has been consciously aware of the target that matches any given response.

Under these conditions we hypothesized four potential outcomes, which are displayed schematically in
Figure 1. Under the null hypothesis of no AC, the quality is low and does not depend upon the intensity
of the feedback. If precognition of the feedback is the underlying mechanism, then we would expect a
linear relationship; the more information in the feedback, the higher the quality of the AC. That is, the
more information in the receiver’s future, the more AC in the session. Suppose, however, that the in-
formation is acquired exclusively in “real-time.” That is, none of the information arises from the future
feedback. Since, by definition, there is no dependency on the feedback, we might expect significant
evidence for AC like the one labeled “Real-time.” A combination of both mechanisms is also shown.
For example, if the feedback is not cognitive (i.c., subliminal), there is little information in the feedback
so that the only open channel is real-time; whereas, the opposite might be true for large amounts of

feedback information.

High
Real—time and

2 Subliminal Precognition Precognition
5 Threshold
'
3
<
2 Real—time
Go]
=]
oO
io, Null Hypothesis
< No AC

5
Z

Intensity of Feedback ——>

Figure 1. Idealized Relationships Between AC Quality and Intensity of Feedback

In this discussion, we have assumed that the feedback content is proportional to the cognitive awareness
of the visual feedback in our proposed experimental outcomes. Based on this assumption, we used the
amount of information available (i.e., the visual intensity) as feedback as the independent variable in an
otherwise straight forward AC experiment.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 3

Approved FOLRRIGRSR ZE99/08108.iSCIA-RDP96-00789RGG32002¢00811404

Experiment Details

In this section, we describe the methodological and technical details of an experiment to determine the
degree to which the quality of the AC depends upon the intensity of the visual feedback.

Tachistoscope Calibration

The crucial independent variable is the amount of visual feedback displayed to the receiver. The magni-
tude of the stimulus is directly proportional to the duration of the receiver’s exposure for a given level of
luminance. In a calibration experiment, volunteers were presented with slides and asked to say when
they were aware of the presentation. We manipulated the magnitude of the stimuli from zero to a value
where the volunteer could recognize the gestalt of a scene. Each stimulus slide was presented for 50
milliseconds (ms), and the magnitude of the stimulus information was adjusted by attenuating the lumi-
nance of the slides over a range of two logarithmic units. In adjusting the magnitude, we relied upon
Bloch’s Law, which says that for presentation time shorter than about 100 ms, the product of time and
intensity is constant (Marks, 1975). Thus, varying the luminance of the stimuli is equivalent to varying
its duration. In addition each slide was appropriately masked with random intensity patterns.

For the luminance calibration, a two-field projection tachistoscope (i.e., Gerbrands G1170) was loaded
with 80 color photographic 35 mm slides (5 opaque and 75 having various luminance contrasts) of natu-
ral and man-made scenes, which where randomly chosen from our larger poo! of 200 photographs from
Natural Geographic magazines (May, Utts, Humphrey, Luke, Frivold, and Trask, 1990). We varied the
juminance contrast of the slides by duplicating them at one of twelve f-stops (including zero) to provide
a target pool having variations in the intensity covering two logarithmic units. The contrast in lumi-
nance for each slide, which may be considered to be the ratio of the brightest to the darkest part of the
slide, was further attenuated in the calibration trials so that some of the slides were above and others
below the observer’s detection threshold. To avoid any possible audio cue to the presentations, the indi-
viduals were presented with white noise in a large, comfortable head phone set, and the tachistoscope
was located in an adjacent room.

The calibration slides were tachistoscopically back-projected onto a 14-inch-square frosted glass win-
dow. Figure 2 shows the floor plan that was used both for the calibration and the AC measurements.
During the calibration experiments, the window cover was fully open; however, during the AC sessions,
the cover was positioned to prevent the monitor from seeing the frosted glass screen.

The tachistoscope was programmed to present each calibration stimulus in numerical order for 50 ms,
followed by a 5-second pause during which the next slide was cycled into position. Slides were atte-
nuated by projecting them through a pair of plane polarizers: one fixed and the other variable. The
juminance of the projected image varied as the cosine of the angle between the two polarizers.

Visual detection, however, is not related to luminance alone. For a fixed luminance, scenes with differ-
ent contrasts will be detected with differing probabilities. At the same luminance, for example, a
photograph of a checkerboard will be easier to detect than a photograph of a mountain cabin in a snow
storm. Thus each of the candidate target slides had to be calibrated with human observers in order to
determine the empirical relationship between detectability and luminance.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 4

Approved For-Release 2OQQ0RMM81sCIA-RDP96-00789RO9S20027000 1074

Window Cover
Tachistoscope Frosted Wind o> Receiver
E | y Table
a 4 NA C_ Monitor
Computer y \_ nie and Terminal
rd ——/
Hall <—_——_——— 3.6m ————_~

Figure 2. Schematic Floor Plan.

Two naive female volunteers participated in the calibration. A complete data set was obtained from one
volunteer, and data trends were confirmed by the second volunteer.

The calibration procedures were as follows. The volunteer was seated approximately three feet from
the projection screen, which was positioned at eye level in the wall between the room in which the appa-
ratus was housed and the room in which the volunteer sat (see Figure 2). The volunteer was permitted
to view the screen and the other contents of the room freely for several (i.e., more than five) minutes to
ensure that she adapted to the ambient illumination level. To screen the sound of the tachistoscope, the
volunteer listened to white noise through earphones. The response was registered by a foot switch that
the volunteer pressed to indicate detection of the stimulus slide. Ina typical session, the variable polari-
zer was set at a predetermined value and each of the 80 slides was presented 5 times. Two sessions were
conducted at each polarizer setting, providing ten data points per slide per polarizer setting. An alter-
native procedure was used when the variable polarizer was set near one of the extremes of the exper-
imental range.’ To reduce the tedium, only those slides near the detection threshold were presented.

Each time a new slide was presented, the volunteer reported whether the presentation was detected
(i.e., sensed at all, regardless of target content). Counters recorded whether a particular slide was de-
tected as well as the proportion of slides detected. From these records, a psychometric function was
generated relating the proportion of the time each slide was detected to the contrast in luminance for
that slide. This function, which relates the contrast in luminance for the slide to its detection threshold,
is an index of the the detectability of the geographic scene depicted in the slide. By using this psycho-
metric function, it was possible to specify not only which slides are subliminal (i.e., never detected), but
also how far above or below the detection threshold each slide lies.

Figure 3 shows a sample of the psychometric curves generated from these data. Six of the 80 slides are
shown by plotting the probability of detecting a given slide as a function of the variable polarizer setting.
We generated a set of curves similar to the ones shown in Figure 3 for all target slides.

* With the extreme settings, the volunteer saw nearly all of the slides or very few of them.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 5

FO eg ee ease Rete Nets A-RDP96-00789RW SQ EMM 994

100 i

Z|
2
2 4
Yu
a 60
a -
is}
= 40
2

20 7
0 89 88 87 85 83 80 76 70 60 50 40

Degrees

Figure 3. Degrees of Polarization (Scaled for Equal Luminance Intervals)

Normally, data would be collected from a large sample of individuals in order to arrive at an average
function, but in this experiment, data from two persons were sufficient for several reasons. First, pilot
studies indicated that interperson variability of stimulus slide detection was quite low. Second, to col-
lapse interperson variability even further, we generated a steep psychometric cure by sampling the ab-
scissae coarsely. For example, if we sampled stimulus slide contrast at only two values—zero and 100
percent contrasts—all observers would respond identically, thus eliminating interperson variation if
the observers possessed normal or corrected vision. In this study, we sampled stimulus target contrasts
at intervals that were found in pilot studies to produce low interperson variability. Finally, for the pur-
poses of this study, interperson variability was not significant because it only shifts the psychometric
function along the abscissa by some unknown amount without changing the shape of the function.
Thus, interperson variability could only result in an erroneous estimate of feedback magnitude. While
these errors may influence the intercept of the function relating the dependent variable (i.e. AC perfor-
mance) to feedback magnitude, the slope of the performance versus magnitude of the feedback is inde-
pendent of these errors.

Figure 4 shows an idealized relationship between visual detectability and luminance for this experi-
ment. To lessen the sensitivity to individual differences in perceiving the feedback, we required that the
psychometric curve rise rapidly through the 50% detection level. We identified eight levels of feedback
intensity that would be used as the values of the independent variable in the experiment. Two were
subliminal (i.e., 0.0 detection probability), and one of these was completely opaque; one was nominally
at threshold; three were at 25, 50 and 75 percent detection threshold, respectively; and two were above
recognition threshold (i.e., 100 percent). These points are shown as arrows in Figure 4.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 «¢

Approved For Release 2000/08/98, i.CIA-RDP96-00789ROG280R27a0811904

1.00F-

&

8 0.75-

ao)

9°

a

§ 0.50;-—

is)

2

A 0.25--
Subliminal _} Luminance
Threshold —

Figure 4. Ideal Psychometric Function

The following example illustrates the procedure that was used for all target slides to select those that
met the criteria shown in Figure 4. We notice in Figure 3, that one of the slides is 50% detectable when
the polarizer was set at 50°, which is one of the detection bands required as illustrated in Figure 4.

Suppose we wished to include the candidate slide, which was detected 50% of the time when the polari-
zer was set at 87.59, in the 50% detection group. We notice a difference of 37.59 in the polarizer setting
between this candidate and the original slide. To account for this difference, we compute cos? (37.59) =
0.629. When we then reduce the exposure of the candidate slide by this factor, it will also be detected
only 50 percent of the time when the polarizer is set at 50 degrees.

In actual practice, we could only change exposure by an integral number of f-stops, so changes of intensi-
ty were always by factors of 2. Working backward, however, the position of all the target slides on the
curve shown in Figure 4 could be determined. Furthermore, all the slides were grouped into bands as
close as possible to the arrows shown in Figure 4.

Experiment Protocol

_ Forty targets, which were randomly selected from the original 80 used in the calibration, were prepared
into eight intensity groups of five targets each using the calibration data and technique described above.
Each intensity group represented the cognitive awareness that each receiver would experience (on the
average) from the feedback. The top two intensities were sufficient to recognize major features, but
insufficient to discern details. By definition, those below the visual recognition threshold could not be
cognitively sensed, and one group constituted no feedback, whatsoever (i.e., zero luminance).

To attempt to maintain some control over precognitively available “answers,” we arranged that at no
future time would a response be knowingly compared to its intended target. For a complete answer to
exist in the future, three pieces of information are needed: (1) the target, (2) the response, and (3) the
comparison between them. The target system was prepared by individuals who had no access to the
responses, and the AC monitor, the assistant, and the receivers had no access to the targets. Finally, the
analysts were never informed which were the correct results on a trial-by-trial basis. Thus in the experi-
ment and forever after only the receiver has had access to all three pieces of information.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 7

Approved For, Release, 204Q/08(08iC1A-RDP26-00789RGR3RORERGON ao.

The slide tray in the tachistoscope was controlled by a computer (Sun Microsystem 3-160) in such a way
that all participants were blind to target selection during a trial. To avoid cuing, for example, the tray
always began and ended in the zero position. When the computer moved the tray, an independent elec-
trical unit, which could be accessed by the computer, counted the tray steps to assure that the intended
target was displayed at the correct time. After the session, the computer repositioned the tray to zero.

‘Two very experienced receivers (receivers 009 and 177), one moderately experienced receiver (105),
and one novice (137) each contributed 40 trials (five at each of the eight intensity levels) at an average
rate of about five trials per week. All receivers except 009 possessed normal vision; receiver 009 is color
blind. The receivers were all in their thirties, and 009 and 177 were male and 105 and 137 were female.
All receivers were completely informed about the nature of the experiment and were asked to sign a
consent form indicating that they were also aware of its potential risks. The protocol had been ap-
proved by an appropriate Institutional Review Board.

All receivers believed strongly that feedback was not necessary for success, but they were uncertain
about the degree to which feedback might contribute to success.

A random order of intensities of the feedback was determined once by a pseudorandom number gener-
ator, which was based on a shift-register algorithm by Kendell and has been shown to meet the general
criteria for “randomness” (Lewis, 1975), and differently for each receiver prior to the start of the receiv-
er’s first trial. Once the order had been set, the trials cycled through the list of intensities until the 40
trials were complete. The sequence of events for each trial was as follows:

(1) Amonitor anda receiver entered a laboratory that contained a table, two chairs, a computer termi-
nal, and a covered 14-inch-square frosted glass window. The window served as a projection screen
for the tachistoscope in the adjacent laboratory. See Figure 2.

(2) The monitor initiated an automatic target selection program on the terminal.

(3) Using the standard feedback shift-register algorithm, which was seeded from the system clock, the
computer randomly selected (with replacement) a target from within the set of five for the given
intensity, stepped the slide tray to that target, and notified the monitor that the trial could begin.
Because of the closed tachistoscope shutters, no illumination of the slide was present on the frosted
screen, and the time to position the slide tray was always set to the time that was required to reach
the most distant slide, regardless of the target slide’s actual position. Thus, there was no time cuing.

(4) The monitor indicated that the trial should begin. For the next 10-15 minutes, the receiver drew
and/ or wrote responses to the intended target.

(5) At the conclusion of the session, the monitor collected the response, and the receiver opened the
screen cover in such a was to shield the monitor from the feedback material.

(6) When the receiver was ready, he or she pressed a button that initiated a single tachistoscope display
of the target. One and only one, 50-ms display appeared on the translucent window screen. (Elec-
tronics prevented the receiver from receiving more feedback after the first button press.) The mon-
itor was instructed not to discuss the experience with the receiver in any way at any time.

(7) The monitor ended the session, and notified the control program from the computer terminal. Af-
ter the computer had returned the slide tray to zero, then and only then, did the monitor and receiv-
er leave the room. All target data were preserved in a computer file.

Data Analysis

The rank-order analysis was used in this experiment Using cluster analysis, all 200 targets had previous-
ly been assigned to relatively orthogonal clusters of nearly equal number of similar targets. An assistant

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 5

Approved For Release 2000/08/08ts CIA-RDP96-00789RADZLNORAIRG 1984

prepared packages (one for. each receiver) consisting of all the responses randomly ordered. Next, the
assistant computer-generated a list (ordered by target number) of seven targets for each response con-
sisting of the actual target and six decoys (a different set of seven for each response). The decoy clusters
were chosen randomly, but uniquely, from the complete set minus the one from which that actual target
was selected. That is, no two decoys were chosen from the same cluster. Once a cluster was selected, the
decoy was randomly selected from within the cluster.

The response material and the target/decoys set of seven photographs (i.e., one target, six decoys) were
presented to two analysts for judging. The analysts arrived at a consensus to rank order each set of seven
targets for each response in accordance with the best to the worst response/target match. For each re-
ceiver, a sum-of-ranks statistic was computed for the sessions. In addition, the data were plotted as AC
quality (i.e., eight minus the assigned rank) versus the feedback intensity.

Results

Table 1 shows the average rank (<Rank>), effect size (ES), Z-score, and its associated p-value for each
receiver in this experiment.”

Table 1.

Results of the Tachistoscope Feedback Experiment

Receiver | <Rank> ES Z P-Value
T1771 2.600 | 0.700 443 | 4.43x10-°
009 3.275 0.363 2.29 0.011
137 3.975 0.013 0.08 0.468
105 4,550 —0.275 —1.74 0.959
Total ; 3.600 0.200 2.53 0.004

The effects size was computed from:

gs = Re,
Nea.
12
where N is the number of possible ranks, which was seven in this experiment; R, and R, were the ex-
pected and observed average rank. The Z-score was computed from ES x Vm. For the individual

receivers, m = 40, and for the total m = 160.

Receivers 177 and 009 produced independently significant results (1-tailed). We can combine data for
all receivers in many ways, but the most conservative is a binomial calculation assuming an event proba-
bility of 0.05. Two significant trials in four attempts corresponds to an exact p-value of 0.014. The totals
shown in Table 1, were computed from the total sum of ranks. The important point, however is that this
experiment produced strong evidence for AC.

* The numbers in Table 1, differ slightly from the table on page 137 of May (1993). The results here were checked against the
recorded sum of ranks from which the average rank was computed. We believe these results to be correct.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 9

APRIONRE ERdRAIRAS h ADDO ML |A-RDP9I6-00789RARZ20OALARRI 1894

Figure 5 shows AC quality plotted against the intensity of the feedback for the four receivers. The low-
est possible quality was one and the maximum was seven (i.e., 8-rank). Shown also is the least squares
best fit line.

8 T T T TT
4
WOOD OO O Oo o 2 © ° J
d
6rF @odo oo o o o > o cm
& LOO GP OO OO o oo 6
a
5 ak 2> ep DOO o 2 8 © 4
g OOO CO Oo ° ° ° °
2b Wooo fo O ° o ° 4
| BPO O00 o fo © °
a ee 1 rare 4. 4 pe ak i
oO 1000 2000 3000 4000
Luminance

Figure 5. Combined Data: Quality vs Feedback Luminance.

We computed a Spearman’s ¢ correlation for each of the receivers and for the combined data. Table 2
shows the results of these calculations.

Table 2.

Spearman’s Rank Correlations for Quality with Feedback

Receiver Q t. df P-Value
177 0.112 0.695 38 0.245
009 —0.297 —1.917 38 0.969
137 0.065 0.402 38 0.345
105 0.150 0.935 38 0.178

Total 0.004 0.050 158 0.480

To place these correlations in perspective, we provide a power analysis for 40 trials. One-tailed correla-
tions of 0.26 and 0.48 would lead to 50% and 80% chance, respectively, of reaching statistical signifi-
cance (Cohen, 1977).

Figure 6 shows the quality versus feedback luminance for the two independently significant receivers
(i.e., 177 and 009). As seen in Table 2, receiver 177 demonstrated a slightly positive correlation with the
intensity of the feedback, while receiver 009 showed a significantly negative correlation.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 ,,

ARBEQMCE FAGERAIGASE £O BMORRAts CIA-RDP96-00789RO922 6027000 11994

Receiver 177 Receiver 009
3 wrTT Tr T T TrTrT 8 ¥ T T T
0000 0 0 oo 9 oO 0°¢O °° °
mp tr @o 0 oo 0° ° 4 6+ ® ° ° ° 4
- a
°o ¢ ° ° ° °
co
g 4F Ooo oo e 7 4h oOo 9 9° ° 4
° o ¢°
2r 9 2b ° ° ° 4
9090 o 9
0 pra awewet 1 oe iveee 1 m 0 diva da podinis ss coord aciconnsct
o 1000 2000 3000 4000 ° 1000 2000 3000 4000
Luminance
Figure 6. Correlations for the two Significant Receivers.
Discussion

The relationship that is easiest to understand is hypothesis 1 in Figure 1 (i.e., increased performance
with increased feedback intensity). We did not observe any such positive correlation overall or with
either of the significant receivers. Receiver 009 produced a significant negative correlation between
performance. and feedback, and at this time, we have no explanation for this result.

This lack of positive correlation, in conjunction with significant evidence of AC, complicates the inter-
pretation considerably. The most obvious conclusion is that the receivers obtained their data in real
time and not from the feedback. However, because of the conceptual difficulties outlined in the
introduction of this paper, we must consider alternative explanations.

One alternative is that the absolute position of the psychometric curve is important. Feedback might be
related to the cognitive experience but only at large values of luminance contrast. If this is true, then we
might not expect functional dependence of AC quality on the feedback in this particular experiment.

An important alternative, however, is that precognition may be exclusively goal oriented, and thus, we
are faced with the unfalsifiability issue. We might not ever be able to interpret process-oriented experi-
ments if this aspect of precognition is true. Therefore, the question of from what time frame does AC
data originate remains unanswered at the present.

In response to a question from Russell Targ in June 1994, we examined, post hoc, the data whenno feed-
back was presented (i.e., 0.0 intensity). Table 3 shows the effect size for all the trials (ESt), zero-intensi-
ty average rank (<Rank>), zero-intensity effect size (ES), Z-score, and its associated p-value for each
receiver.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 11

Approved For Release 2000/0R/08., CIA-RDP96-00789R09 38 RO 2c180 1.404

Table 3.

Results for Zero Intensity Feedback

Receiver ESt <Rank> ES Z P-Value
17 0.700 76 | 0700 | 156 | 0.059 |
009 0.363 3.00 0.500 1.12 0.132
137 0.013 600 | 1.000 | —2.24 0.987
105 | —0.275 5.00 | -0500 | —1.12 0.868

Total 0.200 415 | -0.075 | —0.34 0.631

From these results it appears that the most experienced receivers do not require feedback on a trail-by-
trial basis; receivers 177 and 009 maintain or increase the effect sizes from their sets as a whole. The less
experienced receivers, however, appear to perform considerably below their overall contributions. As
Targ and Targ (1986) also found, trial-by-trial feedback is nor necessary for significant AC.

One of us (May) has been exploring the nature of precognition and suggests that precognition experi-
ments are relatively easy to construct and nearly impossible to understand. For example, if precognition
is goal-oriented (i.e., individuals can “peek” into the answer book), then process-oriented experiments
are difficult, if not impossible, to interpret. It is always possible to include ail the complexity of any given
experiment into a “black box” and consider the final result as the “answer book.” Thus the intervening

complexities are simply not apropos.

One obvious problem with this perspective is that goal-oriented precognition is not falsifiable—an un-
acceptable circumstance in science. For any experiment to be valid, there must be a result. No matter
how cleverly the result is “hidden” from various participants in the experiment or how contrived the
protocol, as long as a result will eventually exist, truly goal-oriented precognition is always possible.

The apparent contradiction among the experimental results from Targ and Targ (1986), Radin (1988),
and May et. al’s (1994) computation of Vassy’s results (1986), could be a manifestation of true goal-
orientation. It is only the Radin study that suggests that a prion target probabilities are important. Ra-
din, however, did not consider global goal-orientation in his discussion. In his experiment, a pseudo
random number generator determined which face of a computer-die would be a prion biased. Since the
seed for that decision was taken when “As soon as the cursor is moved at least one pixel (~.3 mm), ...,”
(Radin, 1988, page 195) all Radin, the subject, had to do was to “peek” into his future for the trial and
compute when to move the cursor to select the seed to correspond to a biased face that did nor land face
up. If this were the case, then we could not interpret Radin’s significant precognition results as evidence
for a dependency upon a priori futures.

There is another technical problem concerning precognition. From a physics point of view, the present
does not exist. It is a single geometric point in space-time. (The present is an interval of time with an
infinitesimally short duration.) All other times are either in the past or in the future, and thus the defi-
nition of real-time AC is problematical.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 12

ARREQMEELESSERARASE 29 BMUOBIRR: CIA-RDP96-00789R0032 AO 2KAOD 1964

In an AC experiment, what constitutes the “answer book” in the future? The most direct candidate is
the reporting of the target material to the receiver after the session (i.e., feedback). It is arguable that
providing no feedback to the receiver precludes precognition as an explanation for successful AC ex-
periments. Since there are no examples in physics that are fundamentally anthropocentric, and if in-
formation from a future time is available to anyone, then most certainly it is available to everyone. The
implication is that receivers could obtain information regardless of the feedback parameters—includ-
ing no feedback at all.

Acknowledgements

We would like to thank Beverly Humphrey for her valuable contribution as an analyst and Thane Fri-
vold for creating the tricky computer control code. Thanks also to the tireless effort of the remainder of
the Cognitive Sciences Laboratory for target selection and technical administration.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7 43

ARBESMECL ARE ALARRIE RO PMO RRs CIA-RDP96-00789ROD32 AR ELADD tubs

REFERENCES

Cohen, J. (1977). Statistical Power Analysis for the Behavioral Sciences, Second Edition. Lawrence
Erlbaum Associates, Hillsdale, New Jersey. Page 86.

Feinberg, G. (1975). Precognition—A memory of things future? Quantum Physics and Parapsychology,
Proceedings of an International Conference held in Geneva, Switzerland, August 26-27, 1974.
Laura Oteri, Editor. Parapsychology Foundation, Inc., New York, 54-73.

Honorton, C. and Ferrari, D.C. (1989). ‘Future Telling:’ A meta-analysis of forced-choice precognition
experiments, 1935-1987. Journal of Parapsychology, 53, 281-308.

Lewis, T. G. (1975). Distribution Sampling for Computer Simulation. Lexington, MA: Lexington Books.

May, E. C. (1993). Technology: A mixed blessing for modern psi research. Psi Research Methodology: A
Re-Examination, Proceedings of an International Conference held in Chapel Hill, NC, October
29-30, 1988. Lisette Coly and Joanne D. §. McMahon, Editors. Parapsychology Foundation, Inc.,
New York, 128-148.

May, E. C., Utts, J. M., Humphrey, B. S., Luke, W. L. W, Frivold, T. J., and Trask, V. V. (1990).
Advances in remote-viewing analysis. Journal of Parapsychology, 54, 193-228.

May, E. C., Utts, J. M., Spottiswoode, S. J. and James, C. L. (1994). Applications of decision
augmentation theory. Submitted to the Journal of Parapsychology.

Radin, D.I. (1988). Effects of a priori probability on psi perception: Does precognition predict actual
or probable futures? Journal of Parapsychology, 52, 187-212.

Targ, E. and Targ, R. (1986). Paranormal perception as a function of target probabilities. Journal of
Parapsychology, 50, 17-28.

Vassy, Z. (1986). Complexity dependence in precognition. Journal of Parapsychology, 50, 235-270.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003200270001-7,,