A PROTOTYPE ANALYSIS SYSTEM FOR SPECIAL REMOTE VIEWING TASKS

• Document Type: CREST
• Collection: STARGATE
• Document Number (FOIA) /ESDN (CREST): CIA-RDP96-00789R002200530001-9
• Release Decision: RIFPUB
• Original Classification: K
• Document Page Count: 28
• Document Creation Date: November 4, 2016
• Document Release Date: May 18, 2000
• Sequence Number: 1
• Content Type: REPORT

~-~.
ed For Release 200 /08 : CIA- P96-007898002200530001-9
Fina! Report-T~sk 6.0.3 October 1989
Covering the Period 1 OoTober 1988 to 30 September 1989
A PROTOTYPE ANALYSIS SYSTEM FOR SPECIAL
REMOTE VIEWING TASKS (U)
SRI Project 1291
J
y o op es
Is document consists of 27 pages
~~~ i c5~ ~
OT RELEASABLE TO
FOREIGN NATIONALS
333 Ravenswood Ave. Menlo Park, CA 94025
roved For~el~~~~S~00~'1/08~((9~~'~~S~~tD~-(~~`T~R002200530001-9
 Approved For Release 2000/ ? CIA- 96-007898002200530001-9
ABSTRAG'T (Ln
~tl~ "? =r' `~` ~ We~liave~ eve~'oped a prototype analysis system for remote viewings conducted
The system uses individual viewers performance histories
is o?~ anferest
ia
-~ -
.
rge
against
'` iii `conj%inction with curxent"data to prioritize a set of possible) ~--~ interpretations of the
I-
r
r
r
r
r
r
r
,,, Approved For Release 2000/0 /08 :? #+pr=ftD 96-007898002200530001-9
 Approved For Release 200011d9~.~481~196-007898002200530001-9
(TJ) TABLE OF CONTENTS
ABSTRACT .................................................................ii
LIST OF TABLES ........................................................... iv
LIST OF FIGURES .......................................................... iv
I INTRODUCTION (U) ...............................................1
II METHOD OF APPROACH (U) ......... < ...................... < ..... 2
A. (U) Fuzzy Set Formalism ....................................... 2
B. (U) Prototype Analysis System ................................... 5
C. (U) Partial Application of Analysis System to Existing Target Pool ..... 7
D. (U) General Conclusions ...................................... 12
k
REFERENCES .............................................................. 13
APPENDIX A .......................................................... .. 14
APPENDIX B ............................................................... 15
UNCLASSIFIED
Approved For Release 2000/08/08 :CIA-RDP96-007898002200530001-9
 r
r
i
i
i
Approved For Release 2000/(~8/08 : CIA-R~jP96-007898002200530001-9
(Ln LIST OF TABLES
1.
(U) Numerical Listing of Targets .............. ........................... 8
---_ __
................ 1.1
_ -.
2.
(U) Technology Cluster ..................................
__
3.
(U) Principal Elements Contained in the Technology Template .................. 11
(Ln LIST OF TABLES
1. t Cluster Diagram for k --"~ Targets ..................... 10
r
Approved For Release 2000/08/08 :CIA-RDFp96-007898002200530001-9
 Approved For Release 2000/08008 :CIA-RD~96-007898002200530001-9
4
4
I INTRODUCTION (U)
w
H
t
r
r
r
r
r
r
f
i
i
(N) Since 1973, when the investigations of the human information-accessing capability
called remote viewing (RV) first began at SRI International," evaluating the quality of the
information obtained has been a continuing challenge. In order to develop valid evaluation
procedures, two basic questions must be addressed:
(1) What constitutes the target?
(2) What constitutes the response?
_' If the RV task is research-oriented, the targets are known, and therefore can be
precisely defined. In ap,W+~.'A9Yl --oriented tasks, however, the targets are generally unknown
and their descriptions are problematical. In both task domains, RV responses tend to consist of
sketches and written phrases. A method to encode unambiguously this type of "natural
language." is one of the unsolved problems in computer science, and there has been little progress
to date. Thus, a complete definition of an RV response is also problematical.
t An o~+p~f[achtan-oriented RV task poses further problems. I=ligh-quality RV does
.not always provide useful ogp4ca-f,?~ For example, the RV may provide additional support for
information that has been verified from other sources;- but provide no new information. In some
cases, however, an overall low-quality RV may provide key elements that positively influence an
analyst's interpretation.
Another characteristic of current laboratory analysis techn;ques is that they do
not provide an a priori assessment of the RV quality. While this is not` a problem in the
laboratory,
_ ,applications require such evaluation. An FtV analyst cannot provide
ratings from the RV alone; rather, the analyst must provide a priori
probabilities that individual RV-response elements (oF concepts) are present at the target site. IC
remains the responsibility of an analyst to~~determine whether such data are ultimately
? Analysis of laboratory RV has been a major part of the ongoing Cognitive
Sciences Program.z-~ For FY 1989, we focused on the development of a prototype analysis
system t}?tat would provide the needed a priori assessments for u{W I. ca~h'av, tasking
(U) References are at the end of this report.
1
Approved For Release 2000i'08i'Q~8 : CIA-RD 96-0078980022005300-01-9
 Approved For Release 2000/08/08-: CIA-RD~R96-007898002200530001-9
r
r
II METHOD OF APPROACH (Ln
r
The analysis of remote viewing (RV) data in an ~, ~; ta,1,~o,., environment differs
C
L
4
~-
i-
r-
considerably from laboratory analysis. Most often, analysts have incomplete or no information
about the target site and are required to provide a priori assessments of data gathered from RV
sessions. In this section we outline a prototype analysis system for -~ RV that uses
concepts From fuzzy set theory, historical archival. data, and "templates" of typical i '-'
targets. In addition, we apply this prototype system to an existing target pool as an illustration of
the power of the technique.
A. (U) Fuzzy Set Formalism
Amore complete description of the full fuzzy set formalism can be found in our
literature.6~' For the purpose of this report, we have summarized that formalism in general terms
that are not specific to either laboratory experiments or C~FpI;Ca~~ tasking.
1. (U) Construction of Target and Response Fuzzy Sets
(U) A formal definition of a target and its associated RV response (i.e., the data
obtained from an RV session) is necessary to any analysis system. To use the fuzzy set method, a
universal set of elements is constructed on which target and response descriptions are based.
These elements should contain descriptive aspects of the target material and incorporate items
that typify responses from the intended viewers. This universal set should also be extendible
(i.e., allow for additional items that may arise in the responses). -
(U) In general, the task of an RV analyst is to assign a membership value (?)
between 0 and 1 to each element in the universal set. The numerical value for each element in a
response is assigned by the degree to which the analyst is convinced that the given element is
present in that response. Membership values for target elements are assigned on the basis of the
degree to which the elements contribute to the target description.
L_.
In the laboratory, the targets are known, so that defining a universal set of
elements is comparatively straightforward.6~7 In Qppl~c~ticv~ tasks, however, defining a single
universal set of elements that is appropriate for all operations is difficult. Because the usual
-task is so highly mission-dependent, defining a single universal set of elements that is
customized to that mission becomes easier.
Approved For Release 2000/0~ 8 : CIA-RD~96-007898002200530001-9
 Approved For Release 2000/08/ ? CIA-RDP 007898002200530001-9
r
The C.~i; c,~,-l-rart analyst, as opposed to an RV analyst, should construct such
a list for each mission. While there may be considerable similarities between element lists for
different missions, undoubtedly the lists will require specialization. In Section II-C below, we
show the construction of one element Iist and how it can be applied to a set of 65 I. _.
targets.
2. (U) Analysis of Complete Responses
Once an appropriate universal set of elements has been created, and fuzzy
sets that define the target and the response have been specified, the comparison between them is
straightforward. We have defined accuracy as the percent of the target material that is described
correctly by a response. Likewise, we have defined reliability (of the viewer) as the percent of
the response that is correct.y Although in the laboratory it is required to provide a posterior
probability estimates of the target-response match, in an ~,I;~,-r,1~ setting, this may be less
important. All that is usually necessary is to describe the accuracy and reliability for complete
responses, and for individual target elements of interest. These quantities for the jth sessions are
n
Wk(Rj t I Tl~k
k=1
Tj = n
WkRj,k
(1)
(2)
n
Wk(Rj I I TI)k
k=1
aj = n
WkTj,k
where the sum over k is called the sigma count in fuzzy-set terminology, and is defined as the sum
of the membership values (?) for the elements of the response, the target, or their intersection,
and n is the number of possible elements as defined by the element List. A fuzzy intersection is
defined as the minimum of the intersecting fuzzy set membership values, In this version of the
definitions, we have allowed for the possibility of weighting the membership values, Wti, to
provide mission-defined relevances.
(U) For the above calculation to be meaningful, the membership values for the targets
must. be similar in kind to those for the responses. For most mission-dependent specifications,
this is generally not the case. The target membership values represent the degree to which a
particular element is characteristic of the target, and the response membership values represent
the degree to which the analyst is convinced that the given element is represented in the
response.
Approved For Release 2000/08/0$ :CIA-R~'96-007898002200530001-9
 Approved For Release 2000/q$~'Q~~,~~6-007898002200530001-9
M
(U) Until RV abilities can encompass the recognition of elements as well as their degree of
target characterization, we are required to modify the target fuzzy set. An analyst must decide
upon a threshold above which an element is considered to be completely characteristic of the
target site. In fuzzy set theory, this is called an a-cut: a technique to apply a threshold to the ?
values such that if the original value exceeds it, reassign the value to 1, otherwise set it to 0. In
this way, the analyst's subjectivity can be encoded in the response fuzzy set, and Equations 1 and
2 remain valid.
3. (U) Analysis of an Individual Element
N
^
(U) Equations 1 and 2 can be simplified to provide an accuracy and reliability on an
individual element basis instead for a complete response. For example, let N be the number of
sessions against different targets that exist in a current archive for a specified viewer. Let a be an
element in question (e.g., airport). Then the empirical probability that element a is in the target,
given that the viewer said it was, is given by
R(E) = Nr , r (3)
where Nc is the number of times that the individual was correct, and Nr is the number of times
that element a was mentioned in the response. R(e) is also the reliability of the viewer, for that
specified element.
(U) To compute what chance guessing would be, we must know the occurrence rate
of element a in the N sessions. Let No be the actual number of times element E was contained in
the N targets. Then the chance-guessing empirical probability is given by
4
Ro(e) can also be considered as the guessing reliability (i.e., the reliability that would
be observed if the viewer guessed a during every session). The more R(e) > Ra(e), the more
reliable the individual is for the specified element.
~ (U) The empirical probability that the viewer said element e, given that it was in the
target, is given by
A(E) = Na .
A (e) is also the accuracy of the viewer for that specified element.
(U) As a numerical example,. suppose a single viewer participated in N = 25 sessions.
Let E _ "airport." Further suppose that No = 5 of the targets actually contained an airport.
Approved For Release 2000/081~~A~IA~F~DP96-007898002200530001-9
 i_
i_
I~
I-
i-
I-
I-
i-
I-
r
r
r
r
i '
Approved For Release 2000/08JQ8 : CIA-RD 96-007898002200530001-9
(v>
Then, Ro(airport) = 0.20 is the chance probability (i.e., guessing airport during every session
would only by 20 percent reliable). Assume that the viewer mentioned airport Nr = 6 times and
was correct Nc = 4 times. Then this viewer's reliability for airports is computed as R(airport) _
0.67 > Ro(airport) = 0.20. The viewer's accuracy for airports is computed as A(airport) = NclNo
= 0.80. Thus in this example, we can conclude that this viewer is reasonably accomplished, at
remote viewing an airport.
B. (U) Prototype Analysis System
1 ~ We assume that an ~ _ . an~lyst has constructed amission-dependent
universal set of elements. We further assume that there are a number of competing
interpretations of the target site in question.
1. (U) Target Templates
The first step in our prototype 'analysis system is to define templates (i.e.,
general descriptions of classes of target types) of all competing target interpretations from the
universal set of elements. 1 ?'
1 _ \ Exactly what the templates should represent is entirely dependent upon
what kind of information is sought. Both the underlying universal set of elements and the
templates must be constructed to be rich enough to allow for the encoding of all the information
_ That is, if neither the set of elements nor the templates can meaningfully
represent information about,i .-five -t~irr3e.}~ _~ ~- ~ then it will be unreasonable to consider
asking, ' I'~leva..rt qucs-hbr~5 cebA.d'
the site. Furthermore, a certain
amount of atomization is necessary because such division into small units provides the potential
for interactions within the universal set of elements. If the profile of a IJ ,facility consists of a
single element, the template would be useless unless the response directly stated that particular
element; rather, the profile should be constructed from groups of elemental features
^~.
There are two different ways to generate target templates. The most
straightforward technique is also likely to be the most unreliable, because it relies on the analyst's
judgment of a single target type. With this method, the analyst, who is familiar with the
intelligence problem at hand, simply generates membership values for elements from the
universal set of elements based upon his or her general knowledge.- Given the time and
resources, the best way to generate template membership values is to encode known targets that
are closely related Each template ? is the average value
across targets, and thus is more reliable. If it is known that some targets are more
Approved For Release 2000/08/08 :CIA-RDP96-007898002200530001-9
 Approved For Release 2000/08/08 :CIA-RDF~96-007898002200530001-9
L
G
4
i-
i-
r-
"characteristic" of the target type than others, then a weighted average should be computed. In
symbols,
(4)
where the sums are over the available targets that constitute the template, wk are the target
weights, and the ?~,k are the assigried membership values for target k.
2. (U) Archival Database
~ ~; A critical feature of an analysis system fo~ -'-"' ARV data is that along
with the current RV data to be evaluated, the individual viewer's past performance on an
element-by-element basis must also be included. For example, if a viewer has been relatively
unsuccessful at recognizing ~j facilities, then a~ ~ reference in the current data should not
contribute much in the overall analysis.
T' As ground truth becomes available for each session, a performance database
should be updated for each viewer to reflect the new information This database should be a
fuzzy set whose membership values for each element are the reliabilities computed from
'Equation 3.
3. (U) Optimized Probability List
,The goal of any ~ ~ ARV analysis system is to provide an a priori
prioritized and weighted list of target possibilities that results from a single remote viewing that is
sensitive to the performance history of the viewer. Assuming that a template exists for each of
the possible) -~` ,interpretations, an analyst should adhere to the following protocol:
(1)
Analyze the RV data by assigning a membership value (?) for each element ir~the
universal set of elements. Each ? represents the degree to which the analyst is
convinced that the particular element is included in the response. ~ ,
(2) Construct a crisp set, R~, as an ~-cut of the original response set. By adopting a
threshold- of 0.5, for example, then the resulting crisp set contains only those
elements that the analyst deems most likely as being present in the response.
(3) Construct an effective response set, R~, as R~ = R~ U R., where Ra is the reliability set
drawn from the archival database. -
6
Approved For Release 2000/ 8 : CIA P96-007898002200530001-9
}
 r--~'~j""'rr
Approved For Release 20fD"8 : CIA-RDP96-007898002200530001-9
~v
Final Report- -Objective D, Task 1 December 1986
Covering the Period 1 October 1985 to 30 September 1986
A SUGGESTED REMOTE VIEWING
TRAINING PROCEDURE (U)
Prepared for: ~~
v~~
This document consists of 86 pages.
oved For e~leas~ 2000/08/08 :CIA-R P96-007898 220530001-9
1 333 Ravenswood Avenue Menlo ark, Galitornia 9 025 U.S.A.
(415) 326-6200 Cable: SRI INTL MPK TWX: 910-373-2046
 Approved For Release 2000/08/ ~8, :CIA-RDP~6-007898002200530001-9
L
4
r
r
r
I"
r
r
(5}
(4) Using this effective response set, compute an accuracy and reliability in accordance
with Equations 1 and 2. Then compute afigure-of-merit, Mi, for the jth competing
interpretations as
Mj=ajxrj .
Of course, the accuracy and reliability use the effective response set from step 3
above.
Order the Ms from largest to smallest value. Since the figures-of-merit range in value
from 0 to 1, they can be interpreted as relative probability values for each of the
alternative target possibilities.
By following such a protocol, an analyst can produce a list of target alternatives that is sensitive to
the current remote viewing yet takes into consideration to the individual viewer's archival record.
.Y
C. (U) Partial Application of Analysis System to Existing Target Pool
(U) We have used an existing target pool (developed under a separate program) as a test
bed for the analysis system described above.
1. (U) Criteria for Inclusion in the Target Pool
Targets in this pool have the following characteristics:
? Each target is within an hour and a half automobile drive of SRI International.
? Each target simulates a1' -----'+#site l ~~interest.
-~
? Each target fits generally within one of five functional categories: Production,
Recreation, Scientific, Storage, and Transportation.
?.~ Each target meets a consensus agreement of experienced RV monitors and
analysts about inclusion in the pool.
(U) The pool consists of 65 targets. Initially, they were divided into 13 groups of five
targets each, where each group contained one target from each of five functional categories. By
carefully organizing the targets.in .this way, the maximum possible functional difference of the
targets within each group was ensured. Table 1 shows a numerical listing of these targets.
7
Approved For Release 2000/0$08 : CIA-~P96-007898002200530001-9
 t
Approved For Release 2000/0 8 : CIA-R P96-007898002200530001-9
_ -~cnsformer Station
--mark
_~~i'lice Dish
- --spons Storage
:: ~: al Fleet
c =:~e1 Quarry
=coming Pool
F. =-:ervatory
~ "-.,n
iC ~.--Ding and Recei H o
d
p
~
L
U
~
?
ac
u.
~
2
k
q
~
a
Z
.o
Q
N
01
O
~
~
1A
N
f0
N
CO
N
O
M Q7
Q
i
m
4
V
~
01
m
'~
~
C
i?
~ W
O
ro
~
E
Q~
c ~
4 m
w
~
U
~ o ~
U w w
+"
eP
~
M h
CD
~
r N
N
N N N
C
_0
~r
4
C
O
~
V
LL
~+
.~
L
Q
Approved For Release 2000/08/08 :CIA-RDP96-007898002200530001-9
 ~nnrnwarl Fnr RPIPacP 7000/OR/08 r1~4_Rr)PAC,_f1f17RQRf1f177f1f1~3flflfl'I _9
~+
~ c ~ ~ ..
~ ~- +.
m ~ w a ~ tmn m
m ~ "-'
w t`- ~ U > H O
aoooooooooao
ooooaoooooao
m
~ o
E
V U
4
m
m ~
~ ?+-? ~ .z
~ ~ F0?- Q
ooooaoa
aoooaoao
ooaaoooo
.0
pprove o
 Aaaroved For Release 2000/08/08 :CIA-RDP96-007898002200530001-9
..
~E c?-~ ..
as ~n 'a m o?' m
oaooooo
m
o .~ ~ .~
U ~ O O
O r N M
et tl' ~ 'at
T r r
oooaoooooaoa
m
~o
o ~ ~ ~ ~ 0 8
a a ? ~ I ~ : I I I I
V ~ U ~ V
O ~ m ~ m m m
a a a a a~ d~ a ~ a a
oaoooooooaao
L
d
C
d
L~
pprov
 "--~- -~-~ ~~~ G~.~e.+ a ~nnninQinQ rl~_RI~PA6.0078AR00 00 0001-9
rr..
C
O
C
IL
a
a
~ ~ ~ ~ ~
~- r- .- .-
aNO ono w a~ ago
oooaoaoooaoao
~~
~~
~ ?~ ?~
r r T
aooooooooaoao
~ ~
T r
N ~
r r
ooaaa
ooooaoo
cD
i
4
m
co
a
Approved For Release 2000/08/08
 Approved For Release 20(~ j( ~8 O~S~~~P96-007898002200530001-9
Appendix B
ANALYSTS' GUIDE TO THE UNIVERSAL SET OF ELEMENTS FOR
FUNCTION ('IT)
(This Appendix is completely UNCLASSIFIED)
Approved For Release 20vO1U87v8S~elA=RDP96-007898002200530001-9
 Approved For Release 20(~~~-~DP96-007898002200530001-9
AN ANALYST'S GUIDE TO THE UNIVERSAL SET OF ELEMENTS (U)
A. (TJ) IIntroduction
(U) This appendix is intended to assist an analyst in using the universal set of elements
shown in Appendix A. We developed six levels of elements ranging from relatively abstract
(information poor) to the relatively complex (information rich).
B. (U) Element Levels and Their Use
(U) The task of the analyst is to assign a membership value between 0 and 1 to each
individual element. For targets, a numerical value will be assigned on the basis of the presence
or absence of each element in terms of functional importance. For responses, the numerical
value will be assigned on the basis of the degree to which the analyst is convincedkthat the
element is contained in the response.
(U) All subsequent commentary is referenced by the element numbers in Appendix A.
Although each level may contain a number of elements, only those individual elements that may
need explanation are listed below.
1. (U) Element Level-Affiliation
(U) "Affiliation" represents an advanced level of remote viewing functioning.
Although we infrequently a observe this advanced functioning, the data are valuable, and,
therefore, are included. Elements in this level can be assigned membership values by asking the
question, "Who owns the target?" There are only three "affiliation" elements:
(1) Commercial/Private.
(2) Government: Federal, state, or local governmental ownership (e.g., municipal
ufilities), but excluding military.
t
(3) Military: military ownership as separate from the above governmental ownership
(e.g., a Navy submarine).
UNCLASSIFIED
Approved For Release 2000/08/08 :CIA-RDP96-007898002200530001-9
 Approved For Release 2 .~~~tDP96-007898002200530001-9
(U) "Function" also represents an advanced level of remote viewing functioning, and
it may represent the most important information with regard to overall function. Elements are
assigned membership values by asking the question, "What is(are) the primary function(s) of the
target?" There are 14 "function" elements, and a few require further explanation:
(6) Distribution: the primary function is to receive ~ to transmit something (e.g., an
electrical transformer station).
(8) Extraction: as in the extraction of minerals from the ground.
(i i) Reception: the primary function is Qn1y, to receive (e.g., a satellite tracking station).
(13) Refining: the primary function is to refine a raw material into an intermediate or
finished product (e.g., a saw mill).
(16) Transmission: the primary function is QuIK to transmit (e.g., a radio tower).
(U) "Attributes" can be thought of as clarification for the "function" level.
Elements are assigned membership values by asking a question similar to, "If the function of the
target is production, then what is being produced?" There are 20 "attribute" elements, and the
following require further explanation:
(18) Animals: animals may.
(20) Biology: the study of living things in general.
(21) Chemistry: also includes chemicals.
(23} Ecology: symbiotic systems in nature, as in ecological zones (e.g., the Bay Lands
Nature Preserve) .
(24) Energy: energy in a broad sense that also includes radio waves.
(29) Nature/Natural: general natural objects (e.g., plants ~, animals).
(32) Plants: plants ~,,
(33) Space exploration: general, includes all experimentation done in space.
Elements 18 .and 32 are given a membership value if the target/response is specifically oriented to
one item. C>therwise element 29 should be assigned a value.
i
(U) Element Level-Modifiers
(U) "Modifiers" can be thought of as a clarification of the "attributes" level.
Elements are assigned membership values by asking a question similar to, "If the function of the
target is production, and vehicles are being produced, then what kind of vehicles are they?"
There are 36 "modifiers" elements, and only element 66 requires further explanation:
UNCLASSIFIED
Approved For Release 2000/08/08 :CIA-RDP96-007898002200530001-9
 Approved For Release 20~~~.~T~~DP96-007898002200530001-9
(66) Symbiotic: symbiotic relationships not subsumed under natural or ecology (e.g., a
cogeneration plant). '
5. QU) Element Level-ObJects
(U) "Objects" contains specific elements not necessarily related to function.
Elements are assigned membership values on the basis of the presence or absence of each object
in terms of functional importance. There are 47 "objects" elements, and the following require
further explanation:
(77) Catwalk: elevated walkway.
(79) Coastline: used only as coastline of an ocean.
(88) High-Technology Electronics: silicon-based technology.
(95) Port/Harbor: port should be marked as in port of departure (e.g., airport, train
station, seaport).
(116) Water-Bounded: only completely bounded bodies of water (e.g., pool or pond).
(117) Water-Canal: manmade.
(118) Water-Large Expanse: the San Francisco Bay should be marked as a large
expanse.
(119) Water-River: also includes stream.
6. (U) Element Level-General/Abstract Items
(U) This level contains the most abstract elements. There are 31 elements, and the
following require further explanation:
(121) Activity-Active: predominant visually active (e.g., an accelerator is very active
electromagnetically, but would be considered passive, because there is little visual
activity); potential activity is considered as passive.
(122) Activity-Passive: predominant visually passive (e.g., a ballpark is passive most of
the time) . s
(123) Activity-Flowing (Water, Air, etc.}: can be natural (e.g. creek) or manmade.
(128) Ambience-Dangerous: perceived and/or physically dangerous.
(140) Colorful: to be used only if especially characteristic.
(141) Modern: to be used only if especially characteristic.
(142) Odd/Surprising: to be used only if especially characteristic.
(143) pld: to be used only if especially characteristic.
(144) Personnel-Few: i to 10 employees mostly full-time.
(145) Personnel-Many: 10 to 1000 employees mostly full-time.
(146) Personnel-None: no full-time employees, but occasional human attention is
allowed.
(148) Size-Large (University Campus): represents a "campus" size area.
(149) Size-Medium (Building): size of typical single buildings.
(150) Size-Small (Human): typically, the size of a human (i.e., 6 feet)
(151) Dull: to be used only if especially characteristic of the color.
UNCLASSIFIED
Approved For Release 2000/08/08 :CIA-RDP96-007898002200530001-9