Managing the Target Pool Bandwidth: Noise Reduction for Anomalous Cognition Experiments

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Managing the Target Pool Bandwidth V4h 23 March 1994

Managing the Target Pool Bandwidth:
Noise Reduction
for
Anomalous Cognition Experiments

by

Edwin C. May, Ph.D
S. James P. Spottiswoode (Consultant)
and
Christine L. James
Science Applications International Corporation
Cognitive Sciences Laboratory
Menlo Park, CA

Abstract

Lantz, Luke, and May (1994) reported in the first of two studies that experienced receivers from the
Cognitive Sciences Laboratory produced significant evidence for anomalous cognition (AC) of static
targets, but showed little evidence for AC of dynamic targets. This result was surprising—it was directly
opposite to the results that were derived from the ganzfeld database (Bem and Honorton, 1994). In
Lantz, Luke, and May’s experiment, the topics of the dynamic targets were virtually unlimited, whereas
the topics for the static targets were constrained in content, size of cognitive elements, and range of
affect. In a second experiment, Lantz, Luke, and May redesigned the target pools to correct this unbal-
ance and observed significant improvement of AC functioning. We incorporate these findings into a
definition of target pool bandwidth and propose that the proper selection of bandwidth will lead to a
reduction of incorrect information in free-response AC. ,

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Managing the Target Pool Bandwidth V4h 23 March 1994

introduction

Effect sizes from forced-choice experiments are much lower than those from free-response studies. For
example, in precognition (Honorton and Ferrari, 1989) and real-time (Honorton, 1975) forced-choice
experiments, the effect size (i.e., Z/,/n) is 0.02, while in the free-response ganzfeld (Bem and Honor-
ton, 1994), the effect size is 0.159. Even if we consider the ganzfeld response as a “forced-choice”
among four alternatives, the x effect size, which converts 1-in-n into an effective binary choice hitting
rate (Rosenthal and Rubin, 1989 and Rosenthal, 1991), is 0.5123-40.0004 for card guessing and
0.5854+0.0287 for the ganzfeld (¢ = 46.2, df ~ 2x 105 p ~ 0). The large t-score is probably due to the
large number of forced-choice trials (i.c., 210°). Considering that the mean of the forced-choice ef-
fect size is 2.50 smallcr than that of the ganzfeld, however, there is clearly a meaningful difference. One
potential source of noise in forced-choice experiments, particularly when trial-by-trial feedback is giv-
en, is memory of the previous trial and knowledge of the complete set of possibilities. For example,
suppose a receiver is asked to guess if a particular card from a normal deck of playing cards is red or
black. Suppose further that there is some putative information coming either from the card or from the
mind of a sender, and that the receiver is a “good” imager (i.c., can easily picture a brilliant image of a
playing card in her/his mind). The receiver’s task, then, can be reduced to simple signal detection. Yet,
if anomalous cognition (AC)" is not a robust information transfer mechanism, and it appears that it is
not, the “signal” is easily lost among the vibrant internal imagery from the memory of all alternative
playing cards. The resulting effect sizes, therefore, are reduced.

The ganzfeld itself was developed as a somatic-sensory noise reduction procedure (Honorton and
Harper, 1974). Honorton argued that by placing a receiver in a sensory-reduced environment, her/his
reactions to the environment would be sharply reduced, encouraging a commensurate reduction of
noise. Based upon the results of our current work, we argue that a major contributor of noise in any
free-response study is cognitive and arises, in part, because of the target pool design.

One result from the ganzfeld experiments suggests that dynamic targets produce stronger results than
static targets (Bem and Honorton, 1994). Lantz, Luke, and May (1994) attempted to replicate this find-
ing in two lengthy experiments in 1992 and 1993. The first of these explored, in a 2x2 design, the rela-
tionship of sender vs no-sender and static vs dynamic target type on the quality of the AC. Since Lantz,
Luke, and May reported no significant effects or interactions due to the sender condition, we will ignore
that aspect of this first experiment. In the second experiment, they conducted all trials without a sender
and changed the characteristics of the target pool. This paper describes the insights gained from these
two studies which led both to the concept of target pool bandwidth, and to a potential way of reducing
noise in free-response AC.

Summary of the first Anomalous Cognition Experiment - 1992

We begin by summarizing the experiment and pertinent results from a study that was conducted in 1992,
the details of which may be found in Lantz, Luke, and May (1994). In the experiment, a static vs dynam-
ic target condition was included to replicate the findings from the ganzfeld.

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

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1
Managing the Target Pool Bandwidth V4h 23 March 1994

Target Pools -— 1992

For the static targets, Lantz, Luke, and May used a subset of 50 of our traditional National Geographic

magazine collection of photographs (May, Utts, Humphrey, Luke, Frivold, and Trask, 1990). These
targets had the following characteristics:

© Topic homogeneity. The photographs contained outdoor scenes of settlements (e.g., villages, towns,
cities, etc.), water (e.g., coasts, rivers and streams, waterfalls, etc.), and topography (e.g., mountains,
hills, desserts, etc.).

© Size homogeneity. Target elements are all roughly the same size. That is, there are no size surprises
such as an ant in one photograph and the moon in another.

© Affectivity homogeneity. As much as possible, the targets included materials which invoke neutral affec-
This pool is perhaps better characterized by what it does not contain. There are no people, animals,
transportation devices or situations in which one would find these items—and no emotionally arousing

pictures.

The dynamic targets, on the other hand, followed similar lines to those from the ganzfeld studies.
Lantz, Luke, and May digitized and compressed video clips from a variety of popular movies or docu-
mentaries. With the exception of cartoons and sexually-oriented material, the clips could contain virtu-
ally anything. Examples included an indoor motor bike race anda slow panoramic scan of the statues on
Easter island. Almost all of the characteristics of the static target pool were violated. The only common
characteristic was thematic homogeneity within any given dynamic clip; across targets there were no
restrictions on content.

Data Analysis and Results - 1992
For each response, a single analyst conducted a blind ranking of five targets—the intended one and four
decoys—in the usual way. The expected mean-chance rank was three. Effect sized were computed by:

(R. - R,)

ES = ————,
/N2~1
i2
where N is the number of rank possibilities (i.c., five in our case) and R, and R, are the expected and

observed average ranks, respectively. The p-values were computed from Z = ES x \/n, where n is the
number of trials.

Each receiver participated in 20 trials for each target type, regardless of sender condition. Table 1 shows
the average rank, the effect size, and its associated p-value for the static target condition. We see that
the combined data is significant and that two of our most experienced receivers, 9 and 372, produced
independently significant results.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Managing the Target Pool Bandwidth V4h 23 March 1994

Table 1.

Results for Static Targets — 1992 Experiment

Receiver | <Rank> ES p-value
9 2.40 0.424 | 0.034
131 3.10 —0.071 0.653
372 2.40 0.424 0.034
389 2.75 0.177 0.240
518 2.60 0.283 0.119
Totals 2.65 0.247 6.8 x 10-3

Table 2 shows the same data for the dynamic target condition.

Table 2.
Results for Dynamic Targets — 1992 Experiment

Receiver | <Rank> |. ES p-value
9 3.00 0.000 0.500
131 2.50 0.354 0.057
372 3.40 —0.283 0.897
389 3.00 0.000 0.500
518 3.10 —0.071 0.624
] Totals 3.00 0.000 0.500

With the possible exception of receiver 131, AC on the dynamic targets failed to show any evidence of
functioning. The difference between these two target conditions was strongly in favor of the static tar-
gets (X* = 3.050, df = 1, p < 0.081).

Hypothesis Formulation and Discussion - 1992

Static targets being better than dynamic ones is surprising—not only because it fails to support the ganz-
feld result, but also because it suggests the opposite. There are a number of possible contributing fac-
tors for this outcome. They include statistical artifacts, idiosyncrasies of our receivers compared to the
ganzfeld participants, and procedural differences. Another possibility may be that, as in the ganzfeld,
rank-order statistics were used. Since absolute measures of AC are better than relative measures in
process-oriented research, and since the target-type inference was based on relative measures, perhaps
this accounts for some of the result. A full discussion of these points may be found in Lantz, Luke, and
May (1994).

We propose a different explanation: a fundamental difference between the expcriment’s dynamic and
static target pools are, in themselves, a source of noise.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1 .

Managing the Target Pool Bandwidth V4h 23 March 1994

The sources of noise in the forced-choice domain are reasonably understood (i.e., memory in conjunc-
tion with complete knowledge of the target pool elements). A new insight for us was another potential
source of noise in the free-response domain. To understand this noise source, we must first assume that
AC data are weak and difficult to recognize. Target pools which contain a large number of diverse cogni-
tive elements, in conjunction with receivers who believe that this is the case, are a source of noise. Re-
ceivers will tend to report any imagined impressions, since those impressions might be part of the target.
Since AC is assumed to be weak, most of the generated impressions are from the receiver’s imagination
rather than from the target. Furthermore, it follows that the noise will increase when these impressions
are unable to be internally edited and must be reported. Thatis, noise is generated not so much from an
active imagination, but imagination coupled with an agreement not to edit the internal experience.

Editing our internal experience is something we all do in our daily communication: we rarely report to a
friend that our mind momentarily wandered during an interesting discussion. Humans appear to have
an ability for multi-proccssing, but we use situational filters to communicate coherently. So, why would
we deny this same ability to participants in AC experiments? In Figure 1, we represent schematically the
contributions to the noise produced by memory and the noise produced by not editing imagination.

Combination

/
Unedited
Imagination

\
Memory

Noise
Selected Photographs

Unselected Movies

Universe

2 5 ‘ 0
Differentiable Cognitive Elements in the Target Pool

Figure 1. Schematic Representation of Sources of Cognitive Noise

As the number of differentiable cognitive elements in a target pool increases from two (for a binary
choice) to nearly infinite (for the universe), we propose that there is a trade-off between noise arising
from memory and noise arising from unedited imagination. For target pools containing fewer ele-
ments, the noise contribution from memory (i.¢e., the curve labeled “Memory” in Figure 1) exceeds im-
pressions arising from edited imagination. Regardless of one’s internal fantasies, there is usually a com-
plete protocol restriction on allowable responses. The reverse is true for target pools that contain a
large number of cognitive elements: the contribution to the noise because of unedited imagination ex-
ceeds that arising from memory. In this case, protocols usually suggest that receivers report nearly all
internal impressions (e.g., in the ganzfcld protocol), and since there will likely be far more of these im-
pressions than there are target elements, the noise is increased. At the same time, since there are a large
number of elements, and because it is difficult to remember all possible elements and their factorial’

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1
Managing the Target Pool Bandwidth V4h 23 March 1994

combinations, the contribution to the noise due to memory is reduced. We suggest that our National
Geographic magazine target pool represents a good compromise: there are enough differentiable ele-

ments to reduce the effects of memory, but few enough to allow reasonable editing of internal experi-
ences that arise from imagination.

The receivers in our experiments have, over time, learned the natural limitations of the National Geo-
graphic target pool by experience and by instruction. They have become skilled at internal editing and
do not report impressions that they know are absent from the overall target pool—thus there is less
incorrect material in their responses.

In Lantz, Luke, and May’s 1992 experiment, where the dynamic targets could be virtually anything, the
receivers were unable to produce significant evidence of AC. They also produced, what is for us, signifi-
cantly reduced functioning with static targets. We speculate that this drop of functioning in both target
conditions arose because the protocol would not allow the receivers to edit their internal experience.
Since the dynamic targets could be consist of anything, and since they were blind to the static-vs-dynam-
ic target condition, they were unable to edit their imaginations, even for the static targets. To illustrate
this point, suppose that half the target pool were ESP cards and the other half were the ganzfeld dynam-
ic targets, but the receivers were blind to the target condition. In any given trial, even though the target
is actually the star ESP card, the receiver is inclined to report all internal imagery, whether it be cartoon
figures, car races, and/or sex scenes from movies. This increased the incorrect information over what it
would be for a simpler target pool of ESP-cards alone.

A strong word of caution is in order. Editing of internal experience because of sensory knowledge of the
target pool cannot inflate a differential rank-order statistic. It will, however, bias any rating scale to-
ward larger values. This is not a problem if ratings are used in correlational or comparative studies.

We define target pool bandwidth as the number of differentiable cognitive elements in the target pool.
Forced-choice experiments usually represent small bandwidths, video clips usually represent a large
bandwidth, and the National Geographic magazine photographs represent an intermediate bandwidth.
At this time, the definition is qualitative, but we will indicate ways in which it can be made more quanti-
tative. Nonetheless, the target pool bandwidth concept is testable.

The following hypotheses formed the basis of Lantz, Luke, and May’s second study in 1993:

(1) A significant increase of AC will be observed for dynamic targets if the dynamic pool is designed
with an intermediate target pool bandwidth that matches the static pool from the 1992 study.

(2) An increase of AC will be observed for static targets because the receivers will be able to edit their
internal experience.

Summary of the second Anomalous Cognition Experiment - 1993

The details of the 1993 study may also be found in Lantz, Luke, and May (1994). In that study, they
included a static vs dynamic target condition to replicate the findings from the ganzfeld, but dropped
the sender condition: all trials were conducted without a sender.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1
Managing the Target Pool Bandwidth V4h 23 March 1994

Target Pools - 1993

For this experiment, Lantz, Luke, and May redesigned both the static and dynamic targets with the
constraint that they all must conform to the topic, size, and affectivity homogeneity of the original static
targets. Surprisingly enough, they identified a large number of videos that could be edited to produce
50 National Geographic-like segments: an airplane ride through Bryce Canyon in Utah or a scanning
panoramic view of Yosemite Falls. Lantz, Luke, and May selected a single frame from within each dy-
namic target video clip, which was characteristic of the entire clip, to act as its static equivalent.

Thus, they were able to improve the target pools in two ways:

(1) The dynamic pool possessed an intermediate target pool bandwidth.
(2) The bandwidth of the dynamic and static pools were nearly identical, by design.

Data Analysis and Results - 1993

For each response, a single analyst conducted a blind ranking of five targets—the intended one and four
decoys—in the usual way. Lantz, Luke, and May computed effect sizes in the same way as in the 1992
study.

Three receivers individually participated in 10 trials for each target type and a fourth, 372, participated
in 15 trials per target type. Table 3 shows the average rank, the effect size, and its associated p-value for
the static target condition. We see that the combined data is significant and three of the four receivers
produced independently significant results.

Table 3.
Results for Static Targets - 1993 Experiment
Receiver | <Rank> ES p-value
9 2.20 0.565 ~ 0.037

372 1.87 0.801 9.7x 10-4

389 3.10 -0.071 0.589

518 1.90 0.778 72x 1073
Totals 2.22 0.566 | 7.5x10->

Lantz, Luke, and May observed a nearly significant increase of AC for the static targets in the 1993 ex-
periment compared to that of the 1992 experiment (X? = 3.158, df = 1, p<0.075), and three of the four
receivers improved from their 1992 results. Thus, the second hypothesis (ie., an increase in AC for
static targets) was strongly supported. Table 4 shows the same data for the dynamic targets.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Managing the Target Pool Bandwidth V4h 23 March 1994
Table 4.
Results for Dynamic Targets — 1993 Experiment
Receiver | <Rank> ES _p-value
9 1.70 0.919 1.8x 10-5
372 1.93 0.754 1.8x 1073
389 3.00 0.000 0.589
518 2.40 0.424 0.091
Totals 2.22 0.566 7.51075

Using the rank-order statistics above, Lantz, Luke, and May saw no difference between static and dy-
namic targets in their 1993 study. The first hypothesis was confirmed: they observed a significant in-
crease of AC with dynamic targets in 1993 from that of 1992 (X? = 9.942, df= 1,p <1.6x10 3),

A detailed analysis of the static vs dynamic target issue may be found in Lantz, Luke, and May (1994)
and in May, Spottiswoode, and James (1994).

General Discussion and Conclusions

One possible interpretation of the results from Lantz, Luke, and May’s two experiments is that the
noisc was sharply reduced by narrowing the target pool bandwidth. They observed a significant increase
of AC with the dynamic targets and a large increase with the static ones. Caution is advised in that this
analysis is post hoc, and there were a number of potential contributing factors. For example, in the first
experiment, receivers were not monitored and were at distances ranging from a few 100s to 1000s of km
from the targets. In addition, feedback was delayed for a few days due to the delivery time of the U.S.
postal service. In the second experiment, the receivers were monitored, given immediate feedback, and
the targets were meters away. Yet, we find the bandwidth analysis compelling because of its “common
sense” appeal. Since the properties attributed to target pool bandwidth may be subjected to experimen-
tal scrutiny, we urge that such studies be carried out. For example, is there a parabolic-like functional
relationship between the target pool bandwidth and the AC effect size? |

To conduct such experiments, we need to develop a quantitative definition of target pool bandwidth.
This implics a quantitative definition of cognitive content, and we have been applying our fuzzy set
analysis (May, Utts, Humphrey, Luke, Frivold, and Trask, 1990) toward this end. We are also looking at
other measures that might be used. Nonetheless, it seems clear that a quantitative definition of band-
width is within reach. Once realized, and if the target pool bandwidth idea can be verified, we all may
benefit from a specific protocol that will reduce the noise in free-response AC experiments.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1
Managing the Target Poo! Bandwidth V4h 23 March 1994

References

Bem, D. J. and Honorton, C. (1994). Does psi exist? Replicable evidence for an anomalous process of
information transfer. Psychological Bulletin. 115, No. 1, 4-18.

Honorton, C. and Harper, S. (1974). Psi-mediated imagery and ideation in an experimental procedure

for regulating perceptual input. The Journal of the American Society for Psychical Research. 68,
156-168.

Honorton, C. (1975). Error Some Place! Journal of Communication, 103-116.

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

Honorton, C., Berger, R. E., Varvoglis, M. P., Quant, M., Derr, P., Schechter, E. I., and Ferrari, D. C.
(1990) Psi Communication in the ganzfeld. Journal of Parapsychology, 54, 99-139.

Lantz, N. D. and Luke, W. L. W,, and May, E. C. (1994). Target and sender dependencies in anomalous
cognition experiments. Submitted for publication in the Journal of Parapsychology.

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

May, E. C., Spottiswoode, S. J., and James, C. L. (1994b). Shannon entropy as an intrinsic target
property: Toward a reductionist model of anomalous cognition.. Submitted for publication in the
Journal of Parapsychology.

Rosenthal, R. and Rubin, D. B. (1989). Effect size estimation for one-sample multiple-choice-type
data: Design, analysis, and meta-analysis. Psychological Bulletin, 106, 332-337.

Rosenthal, R. (1991). Meta-analytic procedures for social research (Rev. ed). Newbury Park, Ca: Sage.

Schlitz, M. J. and Honorton, C. (1992). Ganzfeld psi performance within an artistically gifted
population. The Journal of the American Society for Psychical Research, 86, No. 2, 83-98.

Approved For Release 2000/08/08 : CIA-RDP96-00789R003000490005-1