POSSIBLE PHOTON REPRODUCTION DURING A REMOTE VIEWING TASK: A REPLICATION EXPERIMENT

• Document Type: CREST
• Collection: STARGATE
• Document Number (FOIA) /ESDN (CREST): CIA-RDP96-00789R002200190001-7
• Release Decision: RIFPUB
• Original Classification: K
• Document Page Count: 23
• Document Creation Date: November 4, 2016
• Document Release Date: May 17, 2000
• Sequence Number: 1
• Content Type: REPORT

2?1:;(1
Final Report--Objective E, Task 1 December 1987
Covering the Period 1 October 1985 to 30 September 1987
POSSIBLE PHOTON PRODUCTION DURING
A REMOTE VIEWING TASK:
A REPLICATION EXPERIMENT
By: G. SCOTT HUBBARD EDWIN C. MAY THANE J. FRIVOLD
PETER J. McNELIS, DSW
CONTRACTING OFFICER'S TECHNICAL REPRESENTATIVE
MURRAY J. BARON, Director
Geoscience and Engineering Center
S+~Cr~R l . 1c 3
ed For Reba 12@6dOM Ar@hagRDlA9'9LO67ftF bin6l 9b 7' U.S.A.
[phone] , Cable: SRI INTL MPK ? TWX: 910-373-2046
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
ABSTRACT
Attempting to verify a claim by the Chinese that light is
emitted in the vicinity of correctly identified remote viewing (RV) target
material, we repeated an experiment first published in FY 1984. In that
earlier experiment, a state-of-the-art, ambient temperature, photon
counting system was used to monitor the target material (35-mm slides of
National Geographic Magazine photographs) . The statistical measure derived
from the photon counting apparatus in that study showed a significant
positive correlation with the RV results (p ~ 0.035). That is, when the
remote viewing was good, there was an increase in the signal detected by
the photon counting system. In addition, we observed two anomalous pulses
having a signal-to-noise ratio of about 20:1 to 40:1. In the present
experiment (FY 1987), we improved all hardware aspects of the previous
work, substantially reducing the background noise level and improving
shielding against artifact. In addition, analysis of the remote viewing
indicates that three out of the four viewers produced independently
significant results. Our analysis of the photomultiplier tube (PMT) data
shows no evidence of any anomalous high count rate pulses, no evidence of
any effect on the PMT output during the RV session, and no evidence of any
significant correlation between RV performance and PMT output. We conclude
that (1) the effect proposed by the Chinese is artifactual in nature.
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
TABLE OF CONTENTS
ABSTRACT ................................................................ ii
LIST OF TABLES AND FIGURES ............................................. iv
I INTRODUCTION ................................................... 1
II METHOD OF APPROACH ............................................ 3
A. Hypotheses and Variables ................................. , 3
B. PMT Hardware ................................................. 3
C. Signal Processing ............................................... 4
D. Experimental Methodology ....I ................................. . 7
E. Control Sessions ........... .. .................................. 8
F. PMT Data .................................................... 8
III RESULTS ......................................................... 10
A. Remote Viewing Results .......................................... 10
B. PMT/RV Correlation Results ........................ . ............. 13
IV DISCUSSION AND CONCLUSIONS ................................... 16
A. PMT Anomalies ................................................ 16
B. RV/PMT Output Correlations ..................................... 17
V ACKNOWLEDGMENTS ............................................. 18
REFERENCES ............................................................... 19
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
LIST OF TABLES
1. Remote Viewing Results for Viewer 009 ...................................... 10
2. Remote Viewing Results for Viewer 105 ...................................... 11
3. Remote Viewing Results for Viewer 177 ...................................... 11
4. Remote Viewing Results for Viewer 807 ...................................... 12
1. Schematic Diagram for the Photon Production Experiment ...................... 6
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
I INTRODUCTION
We have 'conducted an experiment to determine whether there may be detectable
physical effects that can be correlated with remote viewing. *
A remote viewing (RV) experiment consists of four basic elements:
? A target consisting of outdoor scenes, including natural and manmade features.
For convenience, experimental targets are pictures taken from National
Geographic Magazine.
? An individual who intends to describe the target.
? A feedback procedure that displays the target.
? An a priori defined analysis procedure to determine if the target has been
accurately described.
A single trial that encompasses these elements might proceed as follows. Using a
random number generator, one target is chosen from a pool of 112, and the viewer is asked to
describe it without seeing it. After the session, the target photograph is displayed to the viewer as
feedback. The description is analyzed using a Figure of Merit (FM) technique to provide a
statistically meaningful evaluation.', 2t
The experiment described in this report is a follow-on to an FY 1984 program that
investigated possible photon production during a remote viewing task.3 In the FY 1984 program,
we conducted a conceptual replication of work published by the People's Republic of China. The
Chinese claimed that anomalous high rate signals from a photomultiplier tube (PMT) are
observed during sessions in which an RV-like ability known as "exceptional vision" was
successfully employed to identify Chinese language characters concealed in the PMT housing.4, 5
Specifically, we experimentally examined the possibility that light is emitted in the vicinity of
correctly identified remote viewing target material.
During the 1984 study, four viewers were asked to contribute six viewings each. The
targets were 35-mm slides of National Geographic pictures of outdoor scenes concealed within a
PMT housing. Our experiments produced a significant positive correlation (p < 0.035) between
Ir This report constitutes the deliverable for FY 1986, Objective E, Task 1, detailing an
experiment to determine fundamental parameters of feedback, shielding and limits of spatial
resolution for RV.
t
References are listed at the end of this report.
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
the quality of the remote viewing and the output of the PMT and two apparent anomalies (high
count rate pulses) during RV periods. In our discussion of the FY 1984 results, we pointed out
that a follow-on experiment was required, which would remove possible sources of artifact,
reduce the PMT noise, and be conducted with more experienced viewers. We addressed all of
these areas in the present experiment.
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
II METHOD OF APPROACH
A. Hypotheses and Variables
In the absence of light leaks or environmental interference, we wish to determine the
degree to which the output of a PMT focused on the RV target slide is positively correlated with
the quality of RV, as determined by FM analysis. The dependent variable in the RV portion of
this experiment is the overall measure of the remote viewing, i.e., the FM. The dependent
variable for the correlation portion of the experiment is the linear correlation (logistic) coefficient
between the FM and PMT output.
To demonstrate that statistically significant remote viewing had occurred, we required
that the FM for a given session exceed a critical value for which the associated probability is <
0.05. The critical value of the FM was determined from the mean-chance-expectation for each
viewer's session. To claim evidence for a statistical anomaly, we required that the observed linear
correlation coefficient between the FM and the PMT output be significantly different from the
expected lack of correlation. To declare that the experiment had confirmed the hypothesis, we
required that the probability of observing the linear correlation coefficient calculated for the data
from 24 viewings (4 viewers, 6 viewings each) must be p ~ 0.05.
In addition to reexamining the principal result of the FY 1984 experiment, we also
planned to look for possible effects on the PMT output during the RV session, independent of the
FM correlation.
We hypothesized that the PMT output for a given viewer's set of sessions would be
significantly different from the corresponding set of local control trials. To confirm this
hypothesis, a t-test comparing the PMT output for the RV session with that for local control trials
must demonstrate a significant difference (p G 0.05.)
We used the photomultiplier tube light detector system and remote viewing procedure
from the 1984 study, incorporating certain improvements as described below. The set of 112
35-mm slides of National Geographic Magazine sites was again used as our target material.
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
The PMT housing and slide holder were light-tight and constructed of metal that was
grounded and shielded against rf, magnetic, and electrostatic fields. Our entire PMT housing was
further enclosed in a standard photographer's film changing bag so that the slide selection
remained blind.
A light-tight slide holder, which could be opened and closed easily, was fabricated and
fitted to the end flange of the PM tube housing. The holder was positioned the slide within
approximately 2.5 cm of the active surface of the tube. This distance is 1.5 cm greater than in the
1984 study because the new cooled tube housing had a quartz window to provide thermal
insulation while allowing optical transparency into the UV.
The PMT was selected to have an active area equal to or greater than the film area of
the slide. In addition, we required the tube to be sufficiently broad band and sensitive so as to
equal or surpass the device used in the first experiments by the Chinese. We also required the
dark count (background) rate to be as low as possible. To achieve this goal, we added cooling so
that the PMT was cooled and the temperature stabilized at -20'C (:J=l'). This modification
reduced the overall dark count noise by a factor of approximately 50 from the earlier study, and
eliminated slow changes in the count rate caused by ambient temperature drift.
All critical pieces of electrical apparatus were isolated from common mode and
differential mode noise. Because the power requirements for the key experimental apparatus
were too great to make battery power a reasonable option, we surveyed the available ac voltage
regulators and transient suppressors. We selected a Topaz power conditioner as having the most
modern design and providing the best specifications for transient suppression (1 kHz to 10 MHz
per IEEE Standard 587), for voltage regulation (+4% to -8% for a +15% to -25% variation about
120 V), and noise rejection (130-dB common mode, 58-dB normal mode at 10 kHz).
Low level signals from the PMT to the preamplifier and preamplifier to amplifier were
more heavily shielded than in FY 1984; the preamplifier was close coupled to the PMT and the
signal from the preamplifier was conducted through semi-rigid coaxial cable, which provides good
electromagnetic interference attenuation up to GHz frequencies.
C. Signal Processing
The output of the PMT was processed and displayed by state-of-the-art
instrumentation used in nuclear radiation spectroscopy. We selected the multichannel scaling
(MCS) mode of signal processing as the most appropriate for our experiment. In this type of data
acquisition, the amplified pulses from the PMT were counted for a specific length of time (dwell
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
time), and the resulting total was stored and displayed in a single bin (channel). This process was
repeated for each of 1024 channels. In this fashion, a histogram was accumulated showing the
tube count rate as a function of time over a single viewing trial (-.17 minutes). We call this
particular time record of PMT output a spectrum.
Because the voltage output of a photomultiplier tube is directly proportional to the
intensity of the incident light source, we elected to set two windows on the PMT signal. One
window displayed the entire voltage range output, which is dominated by numerous small
amplitude background pulses. We designated this window Region I. The window for Region II
was adjusted to show only large voltage pulses. In this fashion, we were able to monitor the
system for two possible outcomes:
? A significant increase in the number of small amplitude pulses.
? An increase in the frequency of relatively rare, large amplitude events.
The original Chinese claim was that an individual having "exceptional vision" (an
ability roughly equivalent to remote viewing) could produce an anomalous signal from the PMT
consisting of mainly large amplitude pulses.
Because the PMT was in total darkness and no light emitting materials were included
in the sample chamber, virtually all background counts were due to thermionic emission at the
photocathode or dynodes.6 A photon striking the photocathode will produce a signal that is
indistinguishable from that resulting from thermionic emission. Therefore, one cannot say
conclusively whether a statistically significant increase in count rate (above background) is caused
by enhanced thermionic emission or photon production. For simplicity, we have referred to the
putative effect in this report as "photon production," and have calculated our results assuming
that photons are striking the photocathode in the PMT.
A multichannel analyzer (MCA) received, sorted, and stored the signals coming from
the two-windows. A third MCA input came from a signal generator that could be triggered by a
microswitch in the adjoining RV room. That switch was used to mark the beginning and end of
data taking in the RV session. Details of the session are contained in the methodology section
below. A schematic diagram of the equipment used is shown in Figure 1.
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
PHOTOMULTIPLIER
TUBE
All Amplitude
Pulses
Session Marker
Pulses
SINGLE
CHANNEL
ANALYZER
SINGLE
CHANNEL
ANALYZER
2
RS-232
INTERFACE
High Amplitude
Pulses
ftI
FIGURE 1 SCHEMATIC DIAGRAM FOR PHOTON PRODUCTION EXPERIMENT
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
Following an experimental session, the data collected by the MCA were transferred to
a Sun Microsystems 3/160 computer via an RS-232 interface. In our FY 1984 study, the count
rates during control periods in our two regions of interest were approximately 300/second and
10/second, respectively. By cooling the T to -200C, we were able to reduce the average count
rates in the respective regions of interest to 5/second and 0.7/second. Because a single photon
can produce a count, we were sensitive to a 1o increase of approximately JN photons, where N is
the count rate. This figure would correspond to about 2 to 3 excess counts in Region I and less
than 1 excess count in Region II.
D. Experimental Methodology
The slides that served as .the targets during the session were prepared from a pool of
112 National Geographic Magazine photographs. Each slide was placed in a separate opaque
envelope marked with an identification number. Prior to each session, four slides were selected
from the target pool by a computer generated pseudorandom number generator (PRNG). All
four slide envelopes were placed in the changing bag with the PMT housing, then shuffled. One
envelope was selected in a blind fashion; the slide was removed and then placed into the special
holder covering the PMT. That procedure ensured that the slide selected was unknown to
everyone.
Four viewers were asked to contribute six viewings each. In this experiment, the
personnel consisted of four of the best available viewers participating in ongoing RV programs at
SRI.
The photomultiplier tube and preamplifier, the instrumentation for amplifying and
counting the PMT signals, and the computer were all located in a locked instrumentation room
adjacent to the RV session room. The only instrumentation in the RV room was the microswitch
used to signal the start and finish of an RV data acquisition period.
Prior to the arrival of a viewer, the experimenter selected the four slides using the
PRNG, checked the equipment for proper functioning, shuffled the envelopes inside the PMT
housing bag, and loaded the target slide. In addition, the experimenter recorded a control session
under exactly the same conditions as an experimental session except for the absence of the
viewer. We designated these as "local" control sessions. In the statistical analysis, these control
sessions (taken before and after the RV session) serve as the baseline against which the
experimental session was compared.
After the arrival of the viewer, 3 minutes of data were collected before beginning the
RV session. Each time the, viewer was ready to give a response in the RV session, the
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
experimenter marked that time with a press of a microswitch. Closing the switch sent a brief
series of TTL logic pulses which registered in the MCA memory. After the response, the
experimenter again closed the switch for 1 to 2 seconds. These two bursts of pulses clearly
defined the period of RV effort for future analysis of the PMT output. A series of several such
efforts using corresponding marker pairs was typically generated during each 17-minute viewing.
This procedure was followed twice (i.e., viewings were conducted two at a time). Following the
session, the accumulated data were automatically read out into the computer memory. Once this
process was begun, the experimenter removed the target slide from its holder and displayed it (as
feedback) to the viewer using a slide projector.
Following the departure of the viewer, the experimenter recorded another local
control session, and then transferred all data from the computer hard disk to magnetic tape.
Following the manufacturer's recommendations, all of the principal hardware (PMT, amplifier,
MCA, etc.) remained on continuously.
Prior to any RV data collection, 50 control sessions were recorded with no one present
in the experimental area. After the completion of all RV sessions, another 50 control sessions
were. recorded. We have chosen to call these "global" control sessions to distinguish them from
the data collected before and after each RV session. As discussed elsewhere in this report, these
100 sessions allowed us to examine the parent distribution of dark counts, and provided a good
record of the maximum count rates observed under normal conditions.
As described, the experimenter also recorded a control session under exactly the same
conditions as an experimental session except for the absence of the viewer, before and after the
RV session. These sessions were designated as "local" controls and are the baseline against which
the experimental session was compared.
During a 17-minute RV trial, the typical data recorded from the PMT consist of three
count rate records (we designate these as spectra for convenience. These three spectra are
displayed concurrently. Each spectrum has a common x-axis of 1024 channels. In order to
accommodate a 17-minute trial, the dwell time was set at 1 second/channel. The first spectrum
displayed all pulses from the PMT, regardless of their amplitude, that were detected during each
1-second counting period. The second spectrum displayed only those pulses--detected in the
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
same 1-second interval--whose amplitude exceeded a preset threshold which was adjusted to
eliminate all but the largest pulses. This amplitude discrimination was accomplished using a single
channel analyzer (SCA) whose settings were the same as in the FY 1984 experiment. The
remaining spectrum represented RV session dependent timing markers. For the all-amplitude
pulse case, the average counting rate was about 5 counts per channel (1 second). Average rates
in the high-amplitude spectrum were about 0.7/channel.
With the dark count stability provided by the temperature controlled PMT, it was
possible to collect meaningful baseline data. We accumulated one baseline for each RV session.
Because two viewings were always carried out at each meeting, one control session was taken
before and one after the entire RV period.
Our planned statistical analysis procedure was very similar to that used in 1984: a
linear correlation coefficient was computed between the FM and the number of excess (> ? ?
1.65 o) pulses from the All-Pass and High-Pass PMT data. Many viewers report perceiving RV
data during rest periods as well as ostensible 'effort periods. Accordingly, we examined
correlations between FMs and the entire RV session PMT output, as well as the RV data
acquisition periods indicated by the microswitch closures. Because the average pulse rate during
this experiment was considerably less than in 1984, a t-test comparing the average count rates in
the control trials and RV trials was used to determine whether there was any significant increase in
the count rate regardless of the FM. A second analysis program searched the RV sessions for
unusually high or low pulse rates (as compared with the baseline data).
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
A. Remote Viewing Results
Each RV session was judged using an FM analysis. The FM is defined as the product
of two measures: accuracy and reliability. The accuracy of an RV response is the fraction of the
target material that is described correctly. Reliability is the fraction of the response that is correct.
Tables I through 4 show the RV results for each trial. The session number (9001.cr, for example)
incorporates a code for each viewer as well as the chronological sequence of viewings.
Session
Figure of Merit
p-value
9001.1g
0.5714
0.0238
9002.lg
0.3810
0.1961
9003.lg
0.4444
0.0497
9004.lg
0.3333
0.3650
9005.lg
0.0667
0.9233
9006.lg
0.3556
0.2697
Overall p C 0.0450
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
REMOTE VIEWING RESULTS FOR VIEWER 105
Session
Figure of Merit
p-value
9001,rs
0.4571
0.0412
9002.rs
0.1667
0.3486
9003.rs
0.1600
0.3618
9004.rs
0.3333
0.1039
9005.rs
0.0000
1.0000
9006.rs
0.3810
0.0475
Overall p G 0.0488
Session
Figure of Merit
p-value
9001.hs
0.4444
0.2430
9002.hs
0.1143
0.9579
9003.hs
0.3810
0.2978
9004.hs
0.5000
0.2392
9005.hs
0.5952
0.0677
9006.hs
0.6429
0.0136
Overall P C 0.0385
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
REMOTE VIEWING RESULTS FOR VIEWER 807
Session
Figure of Merit
p-value
9001.cr
0.0000
1.0000
9002.cr
0.3333
0.2267
9003.cr
0.5208
0.0240
9004.cr
0.0833
0.7494
9005.cr
0.3750
0.1321
9006.cr
0.1333
0.5911
Overall p C 0.1895, n.s.
From the FM analysis performed for our FY 1984 experiment, we determined that by
computing the p-value for each FM we could determine an average p for each viewer and for all
sessions combined. The overall probability of obtaining that average p-value was then calculated,
either by an exact method for small numbers of sessions7 or by using the central limit theorem for
greater than 20 sessions.8 In the current analysis, an additional test of significance, the Fisher
Chi-square technique,8 has been added to supplement the probability associated with average
p-value for a given series.
The overall p-values given for each-viewer's series as shown in
Tables 1 through 4 were calculated using the Fisher Chi-square technique.
Averaging all p-values for all sessions yielded p(avg.) = 0.3437. Using
the central-limit theorem, the probability associated with that average
value is p < 0.004. Using the Fisher Chi-square method, a p-value of ^
0.0036 was calculated for all 24 sessions, indicating good agreement
between techniques. We observed that three out of the four viewers
independently produced significant results. Such an outcome is an
extremely rare event. If the probability of success is p ~ 0.05, the
binomial probability of obtaining three out of four successful results is p
G 0.00048. These individual and overall remote viewing results are
substantially better than were achieved in the FY 1984 study.
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
B. PMT/RV Correlation Results
As described, the basic data unit was a 17-minute session during which 1024 channel
records of output were collected simultaneously in three records. Each channel contains the
number of pulses recorded during a one-second counting period. For a given session, one of the
three records contains the All-Pass PMT data and one the High-Pass PMT data. For the RV
sessions, the third data record contains the markers that indicate the beginning and the end of RV
effort periods.
In order that we could characterize the overall behavior of the system, we collected 50
sessions, with no one present, both before and after the entire series of RV effort sessions. These
were designated as the global control sessions. To characterize the behavior of the system at time
of the RV, one local control trial was conducted both before and after each set of two consecutive
RV sessions.
1. Global Control Data
An error in the data storage program caused the data from the 50 global control
trials following the entire series of RV sessions to be lost, leaving only the pre-experiment global
controls. Given the stability of the temperature controlled PMT housing, this set of approximately
50,000 counting periods appears sufficient to characterize the long-term behavior and general
pulse distribution of the system. In addition, bracketing the RV sessions with two local control
trials serves to detect any short term fluctuations in the system behavior. As described, the local
trials served as the statistical baseline for calculating possible effects on the PMT during the RV
session, regardless of the FM correlation.
The number of counts/channel for the All-Pass mode ranges from 0 to 84, with a
mean of 4.615 and a standard deviation of 3.471; the values for the High-Pass mode range from
0 to 15, with a mean of 0.700 and a standard deviation of 0.921. The distribution of the global
control data exhibits a long tail which is undoubtedly due not only to thermionic emission but also
to the relatively rare dark count processes mentioned in the PMT literature.8 Suppression of
these sources of dark noise is possible, but sufficiently difficult and expensive to be unwarranted
at this stage of investigation.
2. Local Control Data
All of the RV trials were conducted in pairs, one about 30 minutes after the
other. Both before and after each set of two RV trials, a local control trial was collected to
characterize the system's short term behavior. We observed statistically large differences between
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
the means of the global and local control runs, but normal aging of the PMT or very slow
temperature drifts across hours of operating time could easily account for this difference.
The principal effects we wish to examine are (1) correlations between RV quality
(FM) and pulse rate, and (2) changes in pulse rate between control and RV sessions. From the
above analysis, it is obvious that the a priori decision to use the local control data as the baseline
to analyze the RV data was correct.
3. RV Session Data
Data were collected for each RV trial that marked the beginning and end of effort
periods during the session so that the. periods of viewing could be isolated in the analysis. Thus,
for each RV session, there are four distinct sets of data: either the entire RV session or just the
effort period, and for each, either the All-Pass or the High-Pass mode.
As in the FY 1984 experiment, we selected the "excess counts" measure as the
quantity most appropriate for our statistical analysis. The excess counts is the number of channels
in a given session data record with a count rate greater than a critical value. That value is
determined from the distribution of counts per channel of the preceding and the following local
control trials. The critical count rate is chosen so that the area from that value to infinity
represents 5% of the area under the local control distribution curve. Since the hypothesized
effect results in increased photon production, the excess counts measure will identify significant
fluctuations in the high count rate events (p G 0.05). In the All-Pass mode, the critical value
ranges from 11 to 14; for the High-Pass mode, the cutoff value was 3 for all RV trials.
Once the critical value had been determined for each RV session, the number of
excess counts was tabulated for each viewer's group of six sessions. We then employed a paired
t-test to compare the RV excess counts value to the appropriate local control trials. Only one set
of data out of 20 showed a significant result; thus, the conclusion is that no significant effect is
present in the excess counts measure overall.
We also examined the possibility that some RV interaction with the PMT may
have produced an anomalous high pulse rate event as claimed by the Chinese. We compared the
maximum count rates observed in any RV session (All-Pass, 48 counts/second; High-Pass, 9
counts/second), with the maximum count rates in the local control trials (All-Pass, 58
counts/second; High-Pass, 14 counts/second). It is obvious from these data that no unusual
interaction occurred during the RV session.
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
4. RV/PMT Correlation Data
Our principal hypothesis was: There is a correlation between increased PMT
count rate and the quality (FM) of the associated RV: Four sets of linear correlations were done
for each of the four viewers (i.e., All-Pass and High-Pass for effort periods and total session).
Because the claim was that the RV would result in an increased count rate, we were able to specify
the direction of the interaction and therefore are justified in conducting a one-tailed analysis. No
evidence of correlation was apparent for any measure. We conclude, therefore, that there is no
significant interaction between the quality of RV and PMT output.
Since the FM is bounded [0, 1] and not normally distributed, we performed a
logistic regression analysis as well. Again, no statistically significant correlations were observed.
For completeness, we reanalyzed the FY 1984 data using a logistic regression. As
reported earlier,3 we see a significant correlation between RV quality and PMT "noise" for the
High-Pass data.
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
IV DISCUSSION AND CONCLUSIONS
We conducted a replication of experimental work published in FY 1984 in which we
examined the possibility that light is emitted in the vicinity of correctly identified RV target
material. In FY 1984, the statistical measure derived from the photon counting apparatus showed
a significant correlation with the RV results (p { 0.035, or p < 0.012 using logistic regression).
That is, when the RV was good, there was an increase in the signal detected by the photon
counting system. When the viewing was less accurate, a smaller signal was detected by the
counting system. Out of 22 viewings, we recorded two that contained a photon counting anomaly
having a signal-to-noise (S/N) ratio of about 20-40:1, far below the 100-1000:1 anomalies
reported in the Chinese literature that led to our original study.
In our FY 1984 experiment we concluded that:
"Since we observed both statistical correlations and two suggestive anomalies, we
have concluded that there is sufficient evidence to justify another set of
experiments. To carry out a more definitive investigation, those experiments
should be conducted with the following improvements:
? Add more-experienced viewers to the initial group.
? Cool and temperature stabilize the PM tube to reduce background noise count
rates.
? Introduce yet more stringent electrical isolation from the environment to further
reduce the possibility of artifacts from electrical transients."
In our present experiment we have satisfied all of the foregoing recommendations and
have improved all aspects of the previous work--including the quality of RV, background noise
level, and shielding against artifact.
A. PMT Anomalies
The most extraordinary original claim of the Chinese researchers was that very
high count rate signals (103, to 105 counts/second) were observed to be correlated with correct
"exceptional vision"--a remote viewing-like ability exhibited by selected Chinese participants).
In both of our replications of that work, we have obtained good statistical evidence of RV.
However, from the analysis of distribution of pulse rates in the current experiment, it is clear that
we have not observed any photon counting anomalies that can be considered extraordinary when
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
compared with long-term control trials, It is true that the RV sessions contained occasional pulse
rates which, when compared to typical noise background, had an S/N ratio of about 10:1 (- 50
counts/second, maximum). Because the variance in the count rates is very small, these pulses are
statistically improbable. However, the discussion of PMT dark count processes demonstrates that
such occasional relatively low-rate events are not unknown or unanticipated. We have
demonstrated that such events also occurred in control periods when no RV sessions were being
conducted. It is our opinion that the higher rate events observed in our earlier experiments and
in the Chinese work were the result of transients in the experimental apparatus, which have now
been suppressed through the use of power conditioning, hardening against electromagnetic
interference, and temperature stabilization.
In 1985, SRI had an opportunity to interview Xu Hong Zhang, one of the
principal experimenters in the original Chinese work. When questioned closely about the actual
experimental protocol and instrumentation he described his technique of drilling a hole in the
photomultiplier tube housing to "let the psi in." Even though the housing was enclosed in a
multilayer opaque cloth bag, the hole obviously completely violated the physical and
electromagnetic integrity of the housing. There can be little remaining doubt that the Chinese
observations were artifactual in nature.
B. RV/PMT Output Correlations
Despite the lack of any anomalous large-scale PMT counting events attributable
to RV, the FY 1984 study produced a statistically significant correlation between FM and PMT
output. That correlation at least conceptually paralled the Chinese claim, and appeared less likely
to be artifactual in origin since the correlation was obtained across data generated by 4 viewers
and 22 RV sessions. In repeating the FY 1984 experiment, our principal assumption was as
follows: If the RV process did in some way interact with the PMT system, then reducing the noise
of the PMT system should dramatically enhance the signal and, therefore, the correlation.
Several techniques were used to reduce the background count rate by a factor of 60 in the
All-Pass mode and at least an order of magnitude in the High-Pass mode. Despite this noise
reduction, and substantially better remote viewing, the RV/PMT correlation was not enhanced.
As the analysis demonstrates, the correlation disappeared entirely.
The most obvious explanation of the difference in outcome is that the FY 1984
correlation results were simply fortuitous.
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
VI ACKNOWLEDGMENTS
The authors thank Dr. Jessica Utts for her assistance in conducting the statistical
analysis and for reading the manuscript.
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
 Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7
REFERENCES
1. Humphrey, B. S., May, E. C., Trask, V. V., and Thomson, M. J., "Remote Viewing
Evaluation Techniques, " Final Report, SRI Project 1291 SRI International, Menlo
Park, California (October, 1986)
2. May, E. C., Humphrey, B. S., Mathews, C., and Puthoff, H. E., "Figure of Merit Analysis
for Free Response Data," Proceedings of the 28th Annual Convention of the
Parapsychological Association, Tufts University, Medford, Massachusetts, pp. 343-354,
(August 12-16, 1985)
3. Hubbard, G. S., May, E. C. and Puthoff, H. E., "Possible Production of Photons During a
Remote Viewing Task: Preliminary Results," Proceedings of the 28th Annual Convention
of the Parapsychological Association, Tufts University, Medford, Massachusetts, pp.
355-372, (August 12-16, 1985)
4. Chinese Academy of Sciences, High Energy Institute, Special Physics Research Team,
"Exceptional Human Body Radiation,"-PSI Research (June 16, 1982)
5. Yonjie, Zhao, and Hongzhang, Xu, "EHBF Radiation: Special Features of the Time
Response," Institute of High Energy Physics, Beijing, People's Republic of China, PSI
Research (December 20, 1982)
6. Knoll, G. F., Radiation Detection and Measurement, John Wiley & Sons, New York
(1979)
7. Mood, Graybill, and Boes, Introduction to Theory of Statistics, McGraw-Hill, New York
(1973)
8. Rosenthal, R., Meta-Analytic Procedures for Social Research, Sage Publications, Beverly
Hills, California (1984)
Approved For Release 2000/08/08 : CIA-RDP96-00789R002200190001-7