SCREENING AND SELECTION OF PERSONNEL: THE PERSONALITY ASSESSMENT SYSTEM

• Document Type: CREST
• Collection: STARGATE
• Document Number (FOIA) /ESDN (CREST): CIA-RDP96-00789R002200040001-3
• Release Decision: RIFPUB
• Original Classification: K
• Document Page Count: 20
• Document Creation Date: November 4, 2016
• Document Release Date: May 17, 2000
• Sequence Number: 1
• Publication Date: December 1, 1986
• Content Type: REPORT

Final Report-
Covering the Period 1 October 1985 to 30 September 1986
SCREENING AND SELECTION OF PERSONNEL:
THE PERSONALITY ASSESSMENT SYSTEM (PAS) (U)
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200040001-3
A^Qved For Release 2000/08/08
 Approved For Release 2000/08/0>'C 9002200040001-3
r
IV METHOD OF APPROACH (U)
A. (U) Analysis Domain
The domain of this meta-analysis includes all
applications and psychoenergetic research conducted at SRI International, or under the auspices
of its subcontractors, from 1 October 1973 to 30 September 1988. A priori declared
demonstrations or other activities that were not under the control of SRI International were not
included in the documentation. All other forms of experimentation were included in SRI
International technical reports, unclassified journals, or publications, and thus were part of this
analysis. This database comprises 117 documents with a total of 5,025 pages.
(U) By definition, there is no file drawer problem in this analysis; all items that met the
above criteria were included regardless of their results. Care was exercised to avoid multiple
entries of the same data.
200~ All psychoenergetic phenomena fall broadly into two classes:
(1) Information Processes-those phenomena that involve a passive transfer of
information (e.g., remote viewing, search),
(2) Causal Processes-those putative phenomena that involve an anomalous
interaction with matter (e.g., remote action).
The psychoenergetic effort has been divided into various categories within these processes. The
various categories within this domain are defined as follows:
(1) Forced-Choice-remote viewing where the targets are drawn from a limited
(and known) set of potential symbols (e.g., the integers 0, 1).
(2) RV-Lab-remote viewing where the targets are drawn from a large set of
potential material (e.g., photographs of natural scenes, natural physical
locations), and the experiments are conducted under strict laboratory
conditions.
(3) V remote viewing where the targets are drawn from specific targets of
interest'
(4) Search-remote viewing where the targets are generally known but their
location is unknown
4
Approved For Release 2000/08 8 78 R002200040001-3
 Approved For Release 2000/08/08 002200040001-3
1-
I-
I-
I-
I-
I-
(U)
. For the purpose of this analysis, all putative causal-process experiments are
considered under the general heading of remote action.
(U) Figure 1 shows a schematic representation of these categories and the total number of
individual trials that were conducted within each category.
Psychoenergetics
26,074
Informational
25,045
Forced Choice
19,715
RV-Lab
1,071
RVJ
106
Search
4,153
FIGURE 1 (U) CATEGORIES AND NUMBER OF TRIALS
The total number of psychoenergetic trials (26,074) was collected in 154
different experiments involving 227 different subjects.' All the data were entered into a
computer database management system (DBMS).
B. (U) Database Management System
1. (U) Database Requirements
(U) One of the main purposes of performing a meta-analysis is to be able to look at
data gathered from multiple studies conducted under a wide variety of circumstances. In order
to collect and store the data in a meaningful way, one must know what kind of data
manipulations will be performed. To evaluate the effect of certain parameters on
psychoenergetic functioning, we needed to focus our attention on the conditions of a wide array
of potentially important variables. As a result, the database design is primarily determined by the
data and provides for the selection of information, by experiment, given parameter
specifications.
(U) The number of subjects does not include the preliminary mass screening partici-
pants. The formal screening participants were, however, included in the analysis.
5
Approved For Release 2000/08 8 789 002200040001-3.
Causal
1,029
 Approved For Release 2000/08/08 ? 6140001W02200040001-3
V RESULTS AND DISCUSSION (U)
(U) The results of the meta-analysis are presented here, where possible, in quantitative
analytic terms, and various interpretations are discussed in detail. In addition, items that cannot
be analyzed are discussed from a qualitative perspective.
(U) The analysis proceeds in a top-down fashion in accordance with the hierarchy shown
in Figure 1.
(U) The overall analysis was conducted from three different perspectives:
(1)
All of the data, regardless of the purported skill of the subjects,
(2)
A subset of the data contributed by an experienced group of viewers, GI
(i.e., long-term, generally accepted expert viewers-002, 009, 131, 372,
414, and 504)
r
(3)
All of the data except for the group GI (i.e., All-G1).
7
Table 2 shows the number of trials n, total z score, p value, and effect size d for informational
and putative causal processes and for the combination of the two.
(U) STATISTICAL RESULTS FOR MAJOR CLASSES
(U) Powers-of-ten are shown in parentheses.
11
Approved For Release 2000/08/08. 9.22000400013
Class
Perspective
n
z
p
d
Psychoenergetics
All
25,449
9.37
3.69 (-21)
0.059
Gi
9,825
6.86
3.46 (-12)
0.069
All-G1
15,624
6.53
3.46 (-11)
0.052
Informational
All
24,450
9.07
5.83 (-20)
0.058
G1
9,702
6.69
1.14 (-11)
0.0.68
All-G1
14,748
6.25
1.96 (-10)
0.052
Causal
All
999
2.42
6.39 (-03)
0.077
G1
123
2.06
1.99 (-02)
0.171
All-G1
876
1.89
2.95 (-02)
0.064
 Approved For Release 2000/08/08
Tlm, A 0 200040001-3
I
T
T
r
(U) The number of trials shown in Table 2 differs slightly from those shown in Figure 1.
A few trials in each category were analyzed from a post-hoc point of view and therefore have not
been included in the formal analysis.
The heterogeneity of effect size within each group for all classes is very large (i.e.,
the chi-squares for within-groups were large). This is to be expected for such a global analysis
and is frequently seen in meta-analyses of psychological data.5 The sources of the within-group
variation include the psychoenergetic skill, level of the subjects and fundamental differences
between psychoenergetic tasks.
The data, regardless of subjects or process, show strongly significant evidence for
psychoenergetic functioning (p < 3.69 x 10-21). Both the informational and putative causal
processes show significant evidence of psychoenergetic functioning, as well.
Since p values are strongly dependent upon the number of trials, the modern
trend in meta-analysis is to consider the trial-independent measure of effect size. From this
point of view, the magnitude of the psychoenergetic functioning appears roughly constant for all
the data shown in Table 2, and, according to Cohen's criteria for the interpretation of effect
size,* corresponds to small effects.6 The method of calculating overall effect size, however,
involves a weighted average (see Table 1) and thus may not provide an accurate picture of the
size of the psychoenergetic functioning within a given category. To obtain more insight into the
nature of the functioning, we must examine the data within each category.
B. (U) Results for Categories Within the Informational Process
m Table 3 shows the number of trials, total z score, p value, and effect size for
categories within the informational process. The data show strongly significant evidence for
psychoenergetic functioning for all categories regardless of subjects. The effect size, however,
begins to demonstrate category differences.
The forced-choice effect size (d = 0.052) is equivalent to the overall effect size
shown in Tab e 2 (d = 0.059). Since the forced-choice category accounts for 77% of the total
number of trials, the effect-size averaging technique biases the overall result. For example, the
effect size (d = 0.209) for the RV-Lab category is significantly larger than for the Forced-Choice
case (X2 = 22.70, v = 1; p ~ 6.63 X 10-1). The RV-Lab effect sizes meet Cohen's criterion for
a medium-sized behavioral effect.
(U) Values of 0. 1, 03. and 0.5 correspond to small, medium, and large effects, respectively.
12
Approved For Release 2000/08/ jR0022000400013
 Approved For Release 2000/08/08 : Cl i 200040001-3
Table 3
(U) STATISTICAL RESULTS FOR INFORMATIONAL CATEGORIES
7
Category
Forced_Choice
it
T
Perspective
p<
19,675
7.42
6.12
(-14)
0.052
9,487
5.82
2.92
(-09)
0.060
10,188
4.69
1.39
(-06)
0.046
966
6.49
4.33
(-11)
0.209
196
5.39
3.49
(-08)
0.385
770
4.55
2.71
(-06)
0.164
9
3.98
3.45
(-05)
1.326
9
3.98
3.45
(-05)
1.326
RV-Search
V
All-G1 3,790 2.61 4.53 (-03) 0.042
G1
' (U) Powers-of-ten are shown in parentheses.
VWWFor the RV-Lab category, the experienced group, G1, performs significantl}.
better than the novice, larger group (X2 = 7.63, v = 1; p G 0.0057).
As in the overall analysis, the data analyzed in Table 3 show a large heterogeneit%
of effect size within each category. The heterogeneity of effect size, however, is significantly
reduced for the experienced subjects in the RV-Lab category. This reduction may result from a
more uniform skill level of the subjects in group G1; this is in general agreement with our
qualitative assessment of their abilities.
Opnly 8.5% of the remote viewing operational trials were analyzed as a formal
experiment. The effect size for these exceeds Cohen's definition of a large effect. The
equirements of emote viewing, however, are less dependent upon the
duality of the viewing than they may be on other factors. Excellent remote viewing does not
necessarily imply good information.
Approved For Release 2000/0/08: 897002200040001-3
 Approved For Release 2000/08/08 : CI 2200040001-3
In the RV-Search category, 91.3% of the data were collected under laboratory
r
The small effect size (d = 0.042) is
commensurate with that found in other laboratories, and may reflect our lack of understanding
about how to elicit this form of psychoenergetic functioning.
I
1
i
T
C. (U) Specific Results for Remote Viewing
(U) In this section we address the specific questions posed in the SOW. In any kind of an
investigation where the general results fall under a statistical regime (i.e., z scores less than about
5), no hard definitions exist for definitive conclusions. The problem is confounded in behavioral
science because many factors, beyond the particular independent variable in question, may
significantly alter the outcome of an experiment. In trying to assess a large body of literature, as
more constraints are placed on the outcomes, fewer within-group trials are available for analysis;
thus, statistical conclusions become more difficult. This is also true for psychoenergetic research.
Yet, it is possible to describe trends, to suggest ways of improving experiments based upon earlier
results, and to obtain clear insights into factors that may affect psychoenergetic functioning.
To ensure the most reliable interpretations of results in what follows below, group
GI has. been used for the quantitative discussion. As was shown in Section V.-B, this group
possessed the most homogeneous set of data for the RV-Lab category and demonstrated a
significant amount of remote viewing ability.
I. (U) Selection/Screening
The selection of individuals who are able to accomplish remote viewing both
in an operational setting and in the laboratory is of paramount importance. As is shown in
Section V.-B., above, group GI provides' the best results for both types of remote viewing.
Throughout the history of the program at SRI, 6 individuals have been able to demonstrate
consistent functioning over a long period of time. This does not mean that, after vigorous
searching, only 6 have been found. Rather, given our
charter for most of
the time period in question, we had little impetus to find other viewers. During fiscal years
1986-1988, it became clear that a greater number of talented viewers was needed for both
applications and research.
Approved For Release 2000/08/08: 9022000400013
 Approved For Release 2000/08/08 : Cl4- j200040001-3
r
r
I
r
r
r
r
Prior to FY 1986, little was known about how to select good viewers. There
was little systematic research either at SRI or within the field in general, and what was available
was inconclusive or contradictory. The effort that began in FY 1986 encompassed a broad
approach to the problem. We initiated three different types of quantitative approaches:
self-report personality tests, neuropsychological testing, and behavioral testing (i.e., the
Personality Assessment System-PAS). In addition, we used one heuristic approach, which
simply asked individuals to try remote viewing.
=The heuristic approach has been quite successful. The efficiency (i.e., the
number of talented viewers found divided by the total number screened) is approximately 1% in
the general population (i.e., groups of self-selected volunteers). Based upon the results of a
mass screening effort, two individuals have been asked to be regular contributors to the project.
One other heuristic source of good viewers is individuals who have noticed a
psychoenergetic ability in their lives. Many viewers in group GI came to the project in this
manner, and a new viewer, claiming similar experiences, was identified in a recent screening
effort. This viewer produced an effect size of 0.440 in 6 remote viewing trials, which contained
many striking qualitative correspondences between targets and responses.
Of the quantitative techniques, the neuropsychological approach was not
successful at predicting performance. The PAS, however, predicted performance of 9 viewers to
a significant degree.
all By far, the best way to select viewers as of this writing is to use individuals
who either have bilities measured in other laboratories, or who have had strong personal
experiences.
One technique not mentioned above holds great promise for the future.
Three individuals from group GI who participated in a neurophysiological study of correlates with
remote viewing produced unusually large central nervous system responses to light stimuli
directed at the eyes. More work is needed to determine if this simple test might be the most
effective way to screen for individuals with excellent remote viewing ability.
15
Approved For Release 2000/08/0 9R 02200040001-3
 ved d-For Release 2000/08/08 CIA-RDP96-00789R002200040001-3
2. (U) Targeting
Targeting is a general term to describe the method by which a viewer is
directed to the intended target. Common techniques that have been employed include the
following:
(1) Beacon-an individual at the site of the intended target.
(2) Coordinates-the geographical orm coordinates of the intended
target.
(3) Abstract-a word or phrase (e.g.,"target") or other abstract
representation of the intended target.
(4) ,$glf-none of the above, the viewer initiates the collection of data.
We examined these techniques in order to determine which provides the best access to a remote
For these four targeting techniques, 183 trials were identified-the
remainder, 13 trials, were listed as "unknown" targeting. The effect size for viewings initiated by
these targeting techniques was 0.401, leading to a p value of 2.92 x 10-8. Thus, there is
significant evidence for remote viewing functioning. The between-groups chi-square is
significant (X2 = 12.58, v = 3; p :C 0.0058), indicating that the effect sizes resulting from these
targeting techniques are not drawn from the same population.
It is difficult, however, to attribute the significant differences to targeting
techniques alone. In none of the experiments could the targeting technique be used as a valid
independent variable, because, in all cases, the viewers and experimenters were not blind to the
targeting condition. Thus, it is possible, even likely, that the viewers' scientific or emotional bias
toward one technique or another confounds the interpretation. Other factors, such as feedback
time and type, or potential physics models of information transfer, also confound the
interpretation.
Given these caveats, beacon targeting appears to provide the best and most
stable results (n = 66, z = 5.305, p < 5.65 x 10-11, d = 0.653).
3. (U) Evaluation and Analysis
(U) The evaluation and analysis of remote viewing data has undergone significant
improvement during our 16 years of investigation. Beginning as a simple blind matching by
judges, the techniques have been improved by the addition of concept analysis (the paraphrasing
of a complex response), discrete descriptor analysis (defining targets and response as the yes/no
answers to a predetermined set of descriptors), and fuzzy set descriptors (defining targets and
responses as fuzzy sets).
.Approved For Release 2000/08/081A4130196-008,9.002200040001-3
 Approved For Release 2000/08/08: LOW 02200040001-3
INN
I-
Ii-
The fuzzy set technique has also been applied to
experiments and found to provide a useful estimate of remote viewing accuracy (the percent of
the intended target that was described correctly) and viewer reliability (the percent of the
viewer's response that was correct).
For rapid evaluation of laboratory experiments, rank-order judging of
~
targets within preselected (i.e., by fuzzy set techniques) target packets is recommended. For
more accurate measures of remote viewing ability, however, the full fuzzy set analysis is
suggested. Determining whether the fuzzy set technique can be applied
is
a topic for further investigation.
4. (U) Training
ftkix training efforts were conducted during the time period under
consideration; three were qualitative and three were quantitative. There is no overall quantitative
evidence that remote viewing can. taught to novice viewers. Of the qualitative efforts, two were
three showed. some qualitative evidence, however, that training improves remote viewing skill.
Quantitative experiments were conducted with 18 novice viewers in three
separate experiments comprising 48 1 trials. In the first group, the novices were self-selected on
the basis of strong interest and previous personal experiences. None had participated in prior
laboratory experiments. The six viewers in this group produced overall significant evidence for
remote viewing (n = 169, z = 1.719, p < 0.043, d = 0.132). None of the viewers, however,
individually or collectively demonstrated significant evidence that training helps a viewer to
improve.
The second group of 9 viewers was selected because the Personality
-an Assessment System predicted that they would exhibit a wide range of remote viewing ability.
Overall, their data did not reach statistical significance (n = 221, z = -0.971, p < 0.834,
d = -0.065). While the best viewer produced an effect size of 0.170, none of the viewers' data
reached statistical significance. None of these viewers individually or collectively demonstrated
significant evidence that training helps a viewer to improve.
In the third group of 3 novice viewers, one demonstrated significant
evidence for improvement (n = 26, z = 3.01, p G 0.0013, d = 0.590).
While significant evidence for remote viewing has been observed, whether
training can improve remote viewing skill has yet to be substantiated quantitatively. It is possible
that knowledge has not yet advanced to the point where we know how to train. Since the data
from viewers in group GI have remained stable over time, we conclude that simple practice does
not appear to improve performance.
17
Approved For Release 2000/08/ 7 9R002200040001-3
 RDP96-00789R0
SRI recommends that investi atio
g
ns mto training be continued. In the
~
meantime, good ~vilewers are more encat., e_--j
5. (U) Role of Feedback
(U)` Feedback is defined as providing the viewer with information about t
target after a remote viewing experiment. Ve he intended
ry few experiments were devised to test the role of
feedback in determining remote viewing .quality. In the early phases of the project, the primary
objective was to provide as good a result as possible, and since feedback appeared not to hind
remote viewing, most of the early sessions always included it in one form or another. er
The strongest evidence about the l
1987 tachistoscope experiment. In that study, subliminal ore medback iiinimals visualdfe by the FY
f
provided to the viewers. Two of the four viewers produced independent evi feedback was
viewing ability (n = 40; z = 2.30, Bence for remote
P < 0.012, d = 0.363, and z = 4.43, p { 4.78 6
d = 0.700, respectively). Neither of these viewers showed any dependenc u on x 10-
the visual feedback, including zero intensity (i.e., no feedback at all). y p the intensity of
The question of the role of feedback
examined feedback time was examined for group G1. We (i.e.,
the time duration after a session before feedback was provided),
and feedback type (e.g., site, false. site, verbal, visual). We found that there were
significant differences among the various feedback times and among the various feedback t! and
back types.
.M To interpret these differences with regard to feedback is difficult. For
example, the significant difference between a 1-hour delay compared to a 5-minute delay ma
result from the fact that most of the 5-minute delay feedback intervals occurr s
in which photographs were used as targets. Since the longer delay occurred in in exnt
used beacons and natural sites as targets, one interpretation is that the observed perime
differenc erenc s that
attributable to target type rather than feedback interval. es are
A similar problem arises in the feedback type category. One clear result.
however, does emerge. The effect sizes for feedback of natural sites (d = 0.734) is si nifican ?
larger than for feedback of the incorrect natural site (d = _ g tl}
Giving false feedback 0.137. X2 = 4.55, v = 1; p ~ 0.0-l2).
appears to inhibit remote viewing.
'Release 2.000/08/08: `iCl
A recent study indicates that feedback in
essential remote viewing experiments is not
.' This result is in qualitative agreement with the findings from our tachistosce
experiment. In forced.-choice experiments, however, Hononon found that the role of feedback
in the precognition experiments was critical.8
789R002200040001-3
 Approved For Release 2000/08/08 : CIA-
While the quantitative results are mixed, viewers indicate that feedback is
psychologically important. We conclude, therefore, that feedback should be provided whenever
possible.
6. (U) Effect of Distance
We examined the effect of distance on the quality of remote viewing.
Distances were divided into four ranges: < 1 km, < 50 km, < 5000 km, and > 5000 km. For the
group G1, there was no effect of distance on the quality of remote viewing (X2 = 3.56, v = 2;
p < 0.167). It is possible to be definitive about this particular result since all confounding
variables tend to increase the chi-square rather than decrease it.
7. (U) Effect of Size of Target
Only one experiment has been conducted that directly addresses this issue.
Photographs were reduced to a spot size of approximately 1 mm in diameter. One viewer from
group G1 produced significant results (n = 6, z = 2.10, p < 0.018, d = 0.857). We are able to
conclude that targets 1 mm in diameter do not inhibit remote viewing quality. No data are
available on targets of varying sizes.
8. (U) Physiological Correlates to Remote Viewing
(U) In the field in general, the search for physiological correlates has not been
successful. Early results indicated that an individual should be moderately relaxed and as free
from physiological stress as possible (e.g., headaches, bathroom demands). These results are not
surprising in that it is likely. that such a "physiological" state would be optimal for any human
activity.
No SRI has examined neurophysiological correlates to remote viewing in two
separate experiments. Specifically, the central nervous system appears to respond to a remote
light flash, and thus provides a correlate to remote viewing. For the two experiments, a total of
four viewers (all from group GI) produced independent significant changes in a-production in
correlation with remote light stimuli.'-10
M
nervous system that respond to remote stimuli be continued. The potential for screening and
training are significant.
Approved For Release 2000/08/08 : CIA R 1284040001-3
SRI recommends that the effort to isolate particular parts of the central
 Approved For Release 2000/08/08 02200040001-3
9. (U) .Psychological Correlates to Remote Viewing
Psychological correlates to remote viewing have provided weak
but
L"
Ill no
r
,
significant, evidence for correlations with some forms of psychological variables. In the early
work with the Personality Assessment System, SRI found that many of the group GI viewers
clustered near each other in PAS space. In later work, the PAS predicted viewer performance to
a significant degree. SRI's work with self-report personality tests has not been successful;
however, Honorton reports small, but significant correlations with the thinking/feeling dimension
in the Myers-Briggs Type Inventory." In general, psychological correlates have been weak
and/or unreliable.
10. (U) Shielding and ELF
The main purpose of searching for shielding against s choenerge
P Y tic
functioning is to provide for a secure environment. I. M. Kogan proposed a model of
psychoenergetic information transfer based on extremely low-frequency (ELF) electromagnetic
radiation.12 In that model, Kogan proposed that the brain is, in effect, a 10-Hz oscillator and
the body is a crude antenna. Radiation at that frequency would exhibit many of the properties of
psychoenergetic functioning known at that time.
am Too few data were collected under known shielding conditions to make
definitive statements with regard to shielding. Two trials were collected in a 30-dB shielding at
10 Hz. These trials showed significant evidence of remote viewing (n = 2, z = 1.92,
p < 0.027, d = 1.358). In another experiment, when the target material was contained in a
SCIF, significant evidence for remote viewing was observed (n = 6, z = 1.91, p < 0.028,
d = 0.780). The trend, however, is clear: electromagnetic shielding does not inhibit
psychoenergetic acquisition of target material.
11. (U) Audio Analysis
.a In a single study involving 6 trials with a single viewer from group G1, a
significant correlation of remote viewing quality with the audio/linguistic character of the
response was found (n = 6, r = 0.995, p < 0.050, d = 0.800). One purpose for determining
within-session correlations with remote viewing quality is to provide for an independent and
a priori measure of quality.
(U) SRI recommends that this type of investigation be continued to determine the
degree to which the result can be generalized across viewers.
Approved For Release 2000/08
 Approved For Release 2000/08/08 Cl
12. (U) Search and Tracking
am As was seen in Section V.-B., above, significant evidence for search was
found overall (n = 3,790, z = 2.61, p G 0.0045, d = 0.042). Most of these trials were collected
in experiments using computer techniques. In a few experiments, however, the target material
was physical objects in a laboratory setting. The effect sizes from these experiments do not differ
significantly from the overall. result.
Search has always been a challenge. On a few occasions, operational use of
search has proved extremely useful data, but on the average, both the laboratory experiments
and operational use have been disappointing. SRI recommends continued effort in search to
determine those factors that can enhance a potentially very useful phenomenon.
13. (U) Precognitive Remote Viewing
ask
d
the first SRI precognition experiment provided significant evidence of the
phenomenon (n = 4, z = 1.73, p C 0.042, d = 0.864).13 From FY 1975 to FY 1987,
precognition was not studied in any systematic manner. During FY 1987, one experiment was,
conducted using natural sites as targets and one of the group GI viewers. The result was not
significant (n = 10, z = -0.476, p < 0.683, d = -0.150). A second experiment using novice
viewers was conducted in the same year. This also did not reach a significant level (n = 55,
z = 0.070, p < 0.472, d = 0.064). Therefore, the results of SRI's investigations are mixed.
However, in a recent meta-analysis of the precognition forced-choice literature conducted by
one of SRI's subcontractors, 50 years of experimentation involving 50,000 subjects showed highly
significant evidence for the phenomenon (n a 106, z = 24.23 , p < 4 x 10-52, d = 0.041). This
result is consistent with the forced-choice real-time studies conducted at SRI (d = 0.052).
Taken as a whole, there appears to be compelling evidence for precognition:
When precognition is used as the underlying assumption for a heuristic model of psychoenergetic
functioning, 15 years of random number generator data fall on the predicted theoretical curve. t'
14. (U) Analytics (Forced-Choice)
c
(U) Forced-choice remote viewing (defined in Section IV.-A.) has traditionally
provided weak but consistent evidence for a psychoenergetic phenomenon. In the experiments
conducted during the Rhine era, over one million trials were conducted with ESP cards (i.e., a
one-in-five target system)." Strong significances were observed, but effect sizes were of the
order of 0.02.
Approved For Release 2000/08/08 2200040001-3
21
 Approved For Release 2000/08/08: CIA-RD 996-~R002200040001-3
I so
ftP Table 3 shows the results for 19,675 trials collected at SRI since 1 v ? t ,*,
effect size is consistent with the early. results of Rhine (d = 0.052). In fiscal years l
`
t1,
p S 0.00015, d = 0.51), meeting Cohen's definition of a strong effect. While
~% as
significant improvement with this viewer during the three years, the number of formal ; ? ,,
small, and thus interpretation is difficult.
SRI recommends that a forced-choice investigation be contt?t;, .
determine if such strong effects can be observed in other viewers.
one of the viewers from group GI was able to increase the effect size by a factor of 10 ?.
1S. (U) Conducting an RV Experiment
No formal experimentation has been conducted to examine
parameters that enhance remote viewing. SRI does not use any formal induction techni,l,,,.
the sessions are conducted in a businesslike atmosphere with the viewer and monit,,,
upright and opposite each other across a table. Since the overall effect size (d = 0.385) 111
for group GI meets Cohen's definition of a medium-sized effect, these session condition, It'll
appear to hinder the phenomenon.
16. (U) Countermeasures
- The first step in investigating countermeasures for remote viewinit. i ? i?
examine whether it is possible to shield against psychoenergetic intrusion. As was discnn?.?.,
Section V.-C.-10, E&M shielding does not appear to be effective.
,To provide an effective shield or a useful physical countermeasure,
be determined whether psychoenergetic phenomena interact with the physical world.
remote action studies conducted at SRI, most of the studies have not demonstrated any ek,,I,
of psychoenergetic interaction with the physical world.
Two exceptions are worthy of discussion. In a study conducted in F1 I , ?,~
involving random number generators, the significant results were consistent with the hi', ,,,,
database of such experiments. Later, it was shown that these results are not due to a hl,~ ~,
interaction, but rather due to precognition. 14
During FY 1975, a striking anomaly was observed when one of the v? ., ,
from group G1 attempted to influence a shielded magnetometer. The device was perturb,-ii ,,
significant manner, but no other experiments were conducted that showed similar non-sum-,,,
results.
22
Approved For Release 2000/08/0 *9689R002200040001-
3
 Approved For Release 2000/08/08 :CIA=f~DP96-00789R 200040001-3
u
v
r
an In an experiment designed to replicate claims made in the People's Republic
of China, 'SRI determined the degree to which pulses from a photomultiplier tube correlated with
the quality of remote viewing. While strong evidence for remote viewing was seen, no significant
correlations with the tube output were observed.
WAt this time, there is no evidence that psychoenergetic phenomena can be
shielded against nor effectively countermeasure d.
23
Approved For Release 2000/08/0 - 002200040001-3
 Appravedfor Release 2000/08/08 - P96-0 8002200040001-3
REFERENCES (U)
1. Zimbardo, P., Psychology and Life, 12th Ed., Scott, Foresman and Company, Boston,
MA, p. 54 (1988) UNCLASSIFIED.
2. Rosenthal, R., Meta-Analytic Procedures for Social Research, Applied Social Research
Methods Series, Vol. 6, Sage Publications, Beverly Hills, CA (1982) UNCLASSIFIED.
3. Light, R. J. and D. B. Pillemer, Summing Up, The Science of Reviewing Research,
Harvard University Press, Cambridge, MA (1984) UNCLASSIFIED.
4. Hedges, L. V. and I. O1kin, Statistical Methods for Meta-Analysis, Academic Press. Inc.,
New York, NY (1985) UNCLASSIFIED.
5. Feingold, A., "Matching for Attractiveness in Romantic Partners and Same-Sex Friends:
A Meta-Analysis and Theoretical Critique," Psych. Bull., Vol. 104, No. 2, pp. 226-235
(1988) UNCLASSIFIED.
6. Cohen, J., Statistical Power Analysis for the Behavioral Sciences, (rev. ed.) Academic
Press, New York, NY (1977) UNCLASSIFIED.
7. Targ, E., R. Targ, and O. Lichtarge, "Realtime Clairvoyance: A Study of Remote Viewing
Without Feedback," The Proceedings of the Parapsychological Association 28th Annual
Convention, Tufts University, Medford, MA, pp. 335-342 (August 1985)
UNCLASSIFIED.
8. Honorton, C., D. B. Ferrari, and G. Hansen, "Meta-Analysis of Forced-Choice
Precognition Experiments," Psychophysical Research Laboratories' Technical Report
(1989) UNCLASSIFIED.
9. May, E. C., R. Targ, and H. E. Puthoff, "Possible EEG Correlates to Remote Stimuli
Under Conditions of Sensory Shielding," Proceedings of Electro77 Professional Program
of the IEEE, Special Session on the State of the Art in Psychic Research, New York
(April 1977) UNCLASSIFIED.
10. May, E. C., W. W. Luke, and T. J. Frivold, "Neurophysiological Correlates to Remote
Viewing (U)," SRI International, Menlo Park, CA, Final Report-Objective D, Task 1,
Project 1291 (December 1988)
11. Private communication.
12. Kogan, I. M., "Is Telepathy Possible?" Radio Eng., Vol. 21, P. 75 (January 1966)
UNCLASSIFIED.
13. Puthoff, H. E. and R. Targ, "A Perceptual Channel for Information Transfer Over
Kilometer Distances: Historical Perspective and Recent Research," Proceedings of the
IEEE, Vol. 64, No. 3, pp. 329-354 (March 1976) UNCLASSIFIED.
14. May, E. C., D. I. Radin, G. S. Hubbard, B. S. Humphrey, and J. M. Utts, "PSI
Experiments with Random Number Generators: An Informational Model," SRI
International, Menlo Park, CA, Final Report Project 8067 (October 1985)
UNCLASSIFIED.
15. Honorton, C., "Error Some Place!," Journal of Communication, pp. 103-116 (Winter
1975) UNCLASSIFIED.
2Approved For Release 2000/08/ 002200040001-3
 Approved For Release 2000/081789 R002200040001-3
(U) Thus, the groups of interest were ranked as ERUa > ERA8 = ERU8 > IRAS
when predicting overall performance and were ranked as the reverse of this when predicting
the significance of the learning within the experiment.
3. (U) Results of Predictions: Correlations Between RV Performance.and PAS
Profiling
(U) Table 5 shows the PAS predictions for overall RV performance as measured
against actual performance--i.e., each trainee's performance as measured by an effect size
estimate (Pearson's r)3 derived from the figure of merit analysis4 p-values. An effect size
estimate is used to normalize for number of sessions.
Actual Viewer
PAS Prediction
Performance
Viewer
Viewer
Effect
Number of
Profile
Prediction
Comments
ID
Size (r)
Sessions (n)
ID
739
ERUa
best
739
0.170
10
210
ERAS
best
137
0.110
23
very
928
ERU8
best
close
in
928
0.082
28
ranking
512
IRA5
best
512 <
-0.131
25
891
IRA5
best
450
-0.139
37
450
IRU4
middle
307
-0.159
25
137
ERU5
middle
might
do well
210
-0.220
23
in IDS,
307
EFU5
worst
but not
in RV
891
-0.267
27
176
EFU6
worst
clearly the
176
-0.279
23
bottom
' PAS prediction coincides with the viewer's actual rank.
Approved For Release 2000/08/ 1111111WO02200040001-3
Table 5
APPAS PREDICTION VS. ACTUAL VIEWER PERFORMANCE
 Approved For Release 2000/08/ 78JR002200040001-3
(U) While the rank order correlation contrasting the top five and bottom four
performers does not achieve statistical significance, it is encouraging that the PAS correctly
identified two out of the three best performers. The failure to identify Viewer 137 is easily
attributed to lack of prior experience with ERUS viewers. Thus, the results tend more to
confirm than to disconfirm the FY 1984 PAS study.
(U) Table 6 shows the PAS predictions for evidence of RV learning as measured
against actual evidence for learning--i.e., each trainee's learning-slope effect size as derived
from the figure of merit slope p-values.
PAS Prediction
Actual Viewer
Performance
Viewer
Profile
Prediction
Viewer
Effect
Number of
ID
ID
Size (r)
Sessions (n)
891
IRAS
Most improvement
739
0.223
10
512
IRA5
928
0.213
28
928
ERU8
137
0.155
23
210
ERA8
210 <
0.082
23
739
ERUa
450
0.046
37
450
IRU4
891
-0.041
27
137
ERU5
176
-0.085
23
307
EFU5
307
-0.392
25
176
EFU6
Least improvement
512
-0.524
25
PASetion coincides with the viewer's actual rank.
(U) The PAS predictions concerning viewer learning are largely unsuccessful.
When the PAS predictions were forwarded to the SRI COTR, however, they were caveated
Approved For Release 2000/08~I0 9R002200040001-3
 Approved For Release 2000/08/08
IV CONCLUSIONS (U)
(U) Several important factors must be noted when assessing the overall efficacy of the
PAS in this study. It is important to observe, for example, that the novice RV training results
are preliminary: final training results are not officially scheduled for delivery until the end of
the first quarter of FY 1987. Although continuation of training with the original nine
participants at this juncture would destroy the double-blind aspect of the PAS study, a
workable solution to this problem has been identified--namely, to continue training with a
new group comprised of the most promising few candidates out of the original group of nine,
augmented with new candidates to whom the monitors and evaluators are blind with respect to
PAS pattern.
The explanation for the observed lack of significance in the
preliminary novice RV training results is presently unknown. One
hypothesis would suggest that the training procedures are simply not
proving effective. This appears unlikely, however, given that significance
was achieved with novice trainees using the same procedures in FY 1984. A
second possibility is that training needs to be of a longer duration.*
This hypothesis can be tested by observing whether significance is achieved
with the selected viewers from this study who continue training.
(U) The PAS results for this study are encouraging and provide a conceptual replication
of the earlier FY 1984 PAS work. In the earlier study, the PAS was used successfully to
predict the top performer out of each of three different training groups. In FY 1986, the
PAS has been used effectively to predict two out of the top three performers in a single
training group. As an empirically driven system, the PAS Reference Groups experience
continual refinement as the PAS data base increases. It is anticipated, therefore, that the
predictive power of the PAS will increase accordingly.
(U) This is more consistent with the apparent indication that aspects of the training results are correlated
with something else, i.e., the PAS.
Approved For Release 2000/08/ 89 T 02200040001-3