CIA-RDP96-00789R002200460001-7

Approved For Release 2000/08/0
-00789R002200460001-7
ř
-L
L
L
Z
L
TABLO
Lr
1
SRI International
C
D
Final Report-
March 1989
Covering the Period 1 October 1988 to 15 February 1989
REVIEW OF THE
PSYCHOENERGETIC
RESEARCH
CONDUCTED AT SRI INTERNATIONAL (1973-1988) (U)
Ti
Speed For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7
International

Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7
UNCLASSIFIED
(U)
TABLE OF CONTENTS
LIST OF TABLES
iv
LIST OF FIGURES
V
I
OBJECTIVE
1
II
EXECUTIVE SUMMARY
2
III
INTRODUCTION
3
IV
METHOD OF APPROACH
4
A.
Analysis Domain
4
B.
Database Management System
5
C.
Statistical Methods
7
V
RESULTS AND DISCUSSION
11
A.
Overall Results
11
B.
Results for Categories Within the Informational Process
12
C.
Specific Results for Remote Viewing
14
REFERENCES
24
-
-
iii
1
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

Approved For Release 20 000LABSIFHED00789R002200460001-7
(U) LIST OF TABLES
1.
Definitions And Meta-analysis Formalism (U)
.
9
2.
Statistical Results For Major Classes (U)
11
3.
Statistical Results For Informational Categories (U)
13
iv
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

Approved For Release 2 dw d:ASSIFIED 00789R002200460001-7
(U) LIST OF FIGURES
1.
Categories And Number Of Trials (U)
5
2
Database Schema Design For Meta-analysis (U)
7
V
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

Approved For Release 2ddddat ASSRPIED00789R002200460001-7
I OBJECTIVE (U)
(U) The objective of Task 6.0.1 of the FY 1989 Statement of Work (SOW) is to assess,
where possible, the experimental results of the research at SRI International since 1973 .*
(U) This report constitutes the deliverable for Statement of Work, Task 6.0.1.
1
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

1
7
-
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7
T
T
II EXECUTIVE SUMMARY (U)
We have conducted a review and analysis of the psychoenergetic research
conducted at SRI International from 1 October 1973 to 30 September 1988. The database
comprises 117 documents with a total of 5,025 pages.
T
T
A total of 25,449 trials were conducted under a variety of protocols. Analysis
indicates that the odds that our results are not due to simple statistical fluctuations alone are
better than 2 X 1020 to 1 (i.e., 2 followed by 20 zeros). Using accepted criteria set forth in the
standard behavioral sciences, we conclude that this constitutes convincing, if not conclusive,
evidence for the existence of psychoenergetic functioning.
The main results are summarized below:
· Remote viewing (RV) can provide useful
information.
. Laboratory and operational remote viewing show the greatest potential for
practical applications.
· Experienced viewers are significantly better than the general population.
. Approximately 1% of the general population possess a natural remote viewing
ability.
· Remote viewing ability does not degrade over time.
. At this time, there is no quantitative evidence to support a training hypothesis.
o Natural scenes are significantly better than symbols as targets for remote
viewing.
· Remote viewing quality is independent of target distance and/or size.
. There is no evidence to support that a psychoenergetic interaction with the
physical world exists.
· Electromagnetic shielding is not effective against psychoenergetic acquisition of
information.
· A potential central nervous system correlate to remote viewing has recently
been identified.
2
Approved For Release 2000/08
P96-00789R002200460001-7
T
T
- LI
-
I

Approved For Release 2099/ Mes : A'S SPFPED789R002200460001-7
III INTRODUCTION (U)
(U) Until recently, the task of assessing any general body of published knowledge was
formidable. Most of the attempts included review articles that were based primarily upon the
informed opinions of the reviewers. It was recognized, however, that in the behavioral sciences
specific problems arose that were unique to those disciplines. For example, many of the
behavioral results are based on a statistical rejection of a null hypothesis, and, using accepted
practices,1* a successful outcome is declared if the odds that the result is not due to a chance
statistical fluctuation are better than 20 to 1. A major problem for reviewers is created when the
behavioral sciences' technical journals refuse to publish results that fail to meet this statistical
criterion. For example, if only one-in-20 studies is published, then the literature may appear to
provide evidence for a phenomenon, but taken with the 19 unpublished studies for every
published one, there is no evidence for a phenomenon. This particular difficulty is called "the
file drawer problem."
(U) This and other problems resulting from the diversity and difficulty of the behavioral
sciences have been addressed in a new review technique known as meta-analysis.2-4
Meta-analytical procedures are most useful when a large number of diverse studies is under
consideration. Meta-analysis provides techniques to clarify the impact of the file drawer
problem and to enable us to combine diverse experiments in a meaningful manner.
(U) The results of SRI's psychoenergetic research encompass a wide variety of
experiments and thus can be addressed with these techniques. The analysis of the SRI data,
however, is simplified because there is no file drawer problem. All experiments that were
conducted have been reported, and thus are included in the analysis.
(U) This report describes the database, the analysis techniques, and the results of 16 years
of psychoenergetic research conducted at SRI International.
(U) References may be found at the end of this report.
3
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

Approved For Release dddd @bASSIFLEET-00789R002200460001-7
2.
(U) Database Design
(U) The database schema that was used consists of four basic tables (people,
documents, experiments, and units), and two basic relationships (author and parameter). See
Figure 2 for an illustration of this schema. The units-table contains information about the lowest
level of statistical analysis in a given experiment. For example, if 6 viewers participated in 20
trials each, the database would contain 6 unit entries-one for the overall result for each viewer.
(U) Although our database management system is a relational database, our
requirements were inherently hierarchical. That is, each of the documents contains several
experiments, and each of our experiments contain several trials. In order to minimize the
redundancy within the database, we attempted to include all pertinent information as high in the
hierarchy as possible. That is, if a parameter or condition applied to an entire experiment, we
would record that data at the experiment level. If, on the other hand, the parameter varied
across units within a given experiment, we made provision to record those data as a function of
unit instead.
(U) The analyses of most of our experiments contain both individual and group
statistics. In order to prevent any trial from being "counted" multiple times, we required that all
experiments be broken up into the "units" which represent the basic grouping of trials upon
which a hypothesis was being tested. Thus, any given trial appears only once in the database yet
we can reconstitute the group statistics at a later time.
(U) This approach offers two advantages. First, any arbitrary parameter which does
not have an explicit slot in the database can be stored, thus providing flexibility. Second, we can
distinguish between "independent variables" and "incidental variables." The former are
variables which are intentionally manipulated by the experimenter, and the latter are actually
parameters which the experimenter either could not control or treated as insignificant.
(U) Some of the documents detail multiple analyses for a given experiment in order
to compare and evaluate standard and new analytic techniques. For this effort, however, we
required that only one analysis be recorded for each experiment, since our primary focus was to
evaluate the parameters that effect psychoenergetic functioning and not to compare different
evaluation techniques. In determining which analysis to enter into the database, we always chose
a blind method over a post hoc method. If a choice still remained, we then always chose the
technique that had been developed first.
6
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

Approved For Release do BbA SSIFDE -00789R002200460001-7
Document
Author
Person
Experiment
Principal Investigator
Subject ID
Unit
Monitor ID
Parameters
One-to-Many
Relationship
Entity
Relationship
UNCLASSIFIED
FIGURE 2
(U) DATABASE SCHEMA DESIGN FOR META-ANALYSIS
(U) The Appendix contains examples of the DBMS input sheets that were used to
encode psychoenergetic data for the database, and the instructions that were given to analysts.
They are included in the Appendix for completeness; there is no further discussion about them in
this report.
C.
(U) Statistical Methods
1. (U) Effect Size Calculations
(U) Effect sizes were calculated for each experiment or condition using the formula
given by Rosenthal: 2
d = In
2
where n is the number of trials and z is the usual normalized output score. If no z score was given
for an experiment, but a p value was, the z that would have given that p value was computed and
used in the formula. The exception to this procedure was for experiments based on a
sum-of-rank statistic. For those, a more appropriate effect size formula was used and is given by
7
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7
(U)
S -
(R+ 1)
d =
R2 - 1
12
2
where S is the average rank and R is the number of choices for each rank.
2. (U) Comparisons Across Classes
(U) Experiments can be categorized in accordance with a number of specific
variables (e.g., type of feedback, type of target, distance between the viewer and the target).
Effect sizes can be examined within a given category and compared across categories. For each
categorization, the following questions are of interest:
(1) Question 1: Is there any evidence of psychoenergetic functioning within
each of the individual categories?
(2) Question 2: Is the level of psychoenergetic functioning constant across
all experiments within a category?
(3) Question 3: Is the level of psychoenergetic functioning constant across
categories?
(4) Question 4: If there are differences across categories, what is the
relative size of the effect in each category?
(U) Table 1 shows the notation that is used in the formalism that answers these
questions.
(U) To answer question 1, compare the average z score in each category with the
standard normal tables.
(U) To answer question 2, compute
k mi
Qw= >>nij (dj-di.) 2.
i = 1 ] = 1
If effect sizes are homogeneous within categories, the distribution of Qw will be approximately X2
with v = (Emi - k) degrees of freedom. The hypothesis of homogeneity is rejected if Qw is large
compared to the chi-square table entry with v degrees of freedom. To test for homogeneity
within a single category, i, compute
Owi = Enij(dij-di.)2 .
j = 1
mi
8
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7
----
-...... ..
---....

.
Approved For Release 2bbp4GbA SA FIE 6 00789R002200460001-7
(U) Similarly, the distribution of Qwi will be approximately X2 with v = (mi - k)
degrees of freedom, and can be examined as above.
Table 1
(U) DEFINITIONS AND META-ANALYSIS FORMALISM
Basic Definitions
k = number of categories
mi = number of experiments in category i; i = 1, ... , k
dij = effect size for experiment j in category i; i = 1, ... , k; j = 1, ... , mi
nij = number of trials in experiment j in category i
Zij = z score for experiment j in category i
Computed Quantities
ni. = > nij = number of trials
j
Enydi
Within Category i
Zi. =
di. = -
1
= average effect size
ni.
Enijzij
1
=
- = di. ni. = average z score
Ini.
n .. = > nį. = total number of trials
1
Across Categories
n ..
ΣΣ
d .. =
= overall average effect size
2 .. = \n. d .. = overall average z score
UNCLASSIFIED
(U) To answer question 3, compute
QB = >ni. (di. - d .. )2.
k
i=1
If effect sizes are homogeneous across categories, the distribution of QB will be approximately X2
with v = k-1 degrees of freedom. Therefore, the hypothesis of homogeneity across categories is
rejected if QB is large compared to the appropriate entry in the chi-square table with v degrees of
freedom.
9
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

Approved For Release 2dddd dd:ASSIFIED00789R002200460001-7
(U) Finally, to answer question 4, approximate 95% confidence intervals may be
computed for the average effect size within a category using
1.96
di. à In ..
Ini.
10
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

Approved For Release
2UNCLASSIFIED00789R002200460001-7
Appendix
CODING SHEETS AND INSTRUCTIONS FOR THE META-ANALYSIS
(This Appendix is UNCLASSIFIED)
25
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

Approved For Release 2UNEI:ASSIFIED00789R002200460001-7
Unit Information
Page:
Coder I.D:
Date:
Form I.D:
Unit
Data
Unit Name
Session
Viewer
Viewer within Condition
Trial
Experiment O:
Unit I.D.
Viewer I.D.
I.D.
Experienced
Novice
Monitor I.D.
Start Date
Date Duration
Start Time
Time Duration
Viewer Location
SRI
Home:
Client:
Field:
0:
Inten-
tional?
Parameters That Differ
Circle or write in all appropriate conditions
YN
Target Name
YN
Targeting Method
Beacon
Abstract
Coordinates
Prompting Self Unknown
O:
YN
Target Type
Ops
Real Site
Photograph
Alpha/Numeric Person
Objects
O:
YN
Target Distance (km)
<1 < 50 < 5000 > 5000
Unknown
0:
Y
Target Location
Inside
Outside
Both
O:
YN
Target: When Selected
Retrocognition
Real Time
Precognition
O:
YN
Shielding Type
Unknown
E&M Cage/Room
Water
SCIF
0:
YN
Feedback Type
None Visual
Audio
Verbal
Intermediate
Site
Unknown O:
YN
Feedback: When
Immediate
<5 min
<1 hr
<1 day
>1 day
Unknown
0
YN
Independent Variable #1
Condition:
YN
Independent Variable #2
Condition:
Statistics
Data
# of Trials
Raw Score
Judgement Score
1
2 3 4
Z-Score
P-Value
Effect Size
Comments:
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7

1
Approved For ReleaseNoBASSIFREE 6-00789R002200460001-7
Publication Information
Page:
Coder I.D:
1
Date:
Form I.D:
Publication Parameters
Data
Title
Authors
SRI Project Number
Document Number
Classification
Total Number of Pages
Type of Report
Final
Mid-year
Interim
Quarterly Progress
Monthly Progress
0:
Date of Publication
Rules for Meta-analysis Coding
Organization
1.
Use one Publication Information sheet for each publication.
2,
Use as few Experiment Information sheets as necessary.
3.
A Unit is the smallest level at which the most basic hypothesis (usually psi versus no psi) was tested.
4.
Results for a hypothesis that cannot be reconstructed from the basic units should be coded as a separate
"experiment". The Type should be listed as O: correlation.
5.
For an experiment, staple together all Unit Information sheets with the Experiment Information sheet on
top. Clip together all experiment packets from the same publication. Number all of the sheets consecu-
tively within a publication.
General:
1.
Circle (or slash) the appropriate choice.
2.
Use [ ] around data to indicate a coder guess or calculation.
3.
If Other (O:) then specify.
Experiment Parameters, Known Target Parameters and Feedback:
1.
Use publication date if Experiment date is unknown.
2.
Generally, independent variables are those manipulated by the experimenter. However, this space can
also be used for variables that differ unintentionally within an experiment. See Rule #4 under "Unit In-
formation".
3.
Example: LANL experiment is coded as follows: Experiment Type: RV-Lab: Principal Hypothesis:
CNS responds to remote, external stimuli; Independent Variable: Timing of remote stimuli.
4.
Targeting Method: Prompting means a sound or gesture (e.g., Gina's bell).
5.
Shielding is for viewer, target, or both.
6.
Most feedback is actually multi-mode. Code the primary mode. Visual Feedback: Photograph (e.g.,
National Geographic Magazine). Audio Feedback: Just a sound (e.g., Bell from the teaching ma-
chine). Verbal Feedback: Verbal debrief (e.g., You did well. The target was ... ). Site Feedback:
Physical visit to the target site (e.g., Outbound experiment).
Basic Analysis:
1,
Rank R = number of choices for ranking, including target and all decoys.
2.
Analysis scale, n = maximum. (e.g., 0 -> 4, n = 4).
3.
Judgment means a qualitative estimate (e.g., by-gosh-by-golly); 1 = complete miss, 4 = complete hit.
Statistic means z-score or F ratio, etc.
4.
Unit Information:
1.
Unit Name is "Session" for a single RV session, but "Trial" for a single forced choice. In forced choice
experiments, there are usually several trials in a single session.
2.
Unit I.D. is blank most of the time. Use Ops tag when appropriate.
3.
Viewer I.D. is according to our most current list. Therefore, if a known viewer was listed under an old
I.D., note the person's name so the current I.D. can be entered in the data base.
4.
Parameters that differ should be filled in only for those cases where "differs" was circled on the Experi-
ment Sheet. If the variable was intentionally manipulated, circle Y.
5.
P-value should be entered as -1 if it is unknown, to avoid confusion with the default missing value code
of 0, which could be a legitimate P-value.
UNCLASSIFIED
Approved For Release 2000/08/08 > CIA-RDP96-00789R002200460001-7

Approved For Release 2dddd : ASS PIED00789R002200460001-7
Experiment Information
Page:
Coder I.D:
Date:
Form I.D:
Sub-experiment or Condition?
y
n
Experiment Parameters
Data
Type
RV-Lab RV-Ops
Forced-Choice
Screening
Training
Search
O:
Date
Pages Within Document
Principal Investigator
Number of Subjects
Principal Hypothesis
Independent Variable (s)
not included below; list
categories or describe
in space provided.
1.
2.
a.
a.
b.
b.
Differs*
Differs"
Experiment Task
Known Target Parameters
Data
Target Name
Targeting Method
Beacon
Abstract
Coordinates
Prompting Self Unknown
Differs" O:
Type
Ops Real Site
Photograph
Alpha/Numeric
Person
Objects
Differs* O:
Distance (km)
<1 < 50 < 5000 > 5000
Unknown
Differs"
0:
Location
Inside
Outside
Both
Differs*
O:
When Selected
Retrocognition
Real Time
Precognition
Differs*
0:
Shielding Type
Unknown
E&M Cage/Room
Water
SCIF Differs*
O:
Feedback
Data
Type
None
Visual Audio
Verbal
Intermediate
Site
Unknown
Differs*
0:
When
Immediate <5 min
<1 hr
<1 day
>1 day
Unknown
Differs"
O:
Basic Analysis
Data
How
Blind
Post Hoc
# of choices
Depndt. Variable
Rating Rank R_
Fuzzy Bit Discr. Bit
Concept
Hits
Scale n
Judgment Match O:
Method
Scott's #rows
FM
Sum-of-Ranks
Statistic
O:
By Whom
SRI
Client
0:
Purpose
RV
PK
Utility
Demonstration
O:
* When "Differs" is circled, information must be entered at the unit level.
UNCLASSIFIED
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200460001-7