CIA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Thruandy afternon
ANALYSIS OF THE SUBJECT-MACHINE RELATIONSHIP
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
TABLE OF CONTENTS
Page
Overview
I.
Statistical Analysis of the Machine Experimental Data
1
Forward-Bakcward State Transition Analysis
7
Experimental Data Randomness Analysis
4
Best Strategy
15
II.
Analysis of S2 Data Responses
16
Strategy of S2
16
Total Color Choices
16
State Transition Color Choice
19
Hit Analysis
21
Learning from Trial to Trial
21
Learning Within a Trial
20
III. Miscellaneous
30
General Trial Information
31
Machine States
31
Plots of Passes
31
Subject - Future Machine State Transition
40
Distribution of Machine State Colors for Machine 2
41
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Page
Figure 1.1.a
Distribution of Machine Yellows Over Trials
6
1.1.b
Distribution of Machine Greens Over Trials
7
1.1.c
Distribution of Blues Over Trials
8
1.1.d
Distribution of Machine Reds Over Trials
9
Figure 1.2
Distribution of Machine Colors When Samples are
Taken Five at a Time
10
Figure 1.3
Distribution of Machine Colors When Samples are
Taken 100 at a Time
11
Figure 1.4
Distribution of Machine Reds When the Samples
are Taken 100 at a Time
12
Figure 1.5
Machine Color Distribution for Machine 1 and
Machine 2 on a Trial to Trial Basis
14
Figure 2.1
S2 Color Choices
17
Figure 2.2
Plot of Number of Hits/Trial
22
Figure 2.3
Frequence Plot of Number of Hits
22
Figure 2.4
Cumulative Success Ratio of Subject (both machines
used)
23
Figure 2.5
Accumulative Probability of Success on Machine 1
24
Figure 2.6
Accumulative Probability of Success on Machine 2
25
Figure 2.7
Hits vs Trial Number for Machine 2
27
. .
Figure 2.8
Total Number of Hits Within a Trial
29
Figure 3.1
Selected Parameter Totals Listed by Trial Number
32
Figure 3.2
Color States of Machine 1 During the Experiment
34
Figure 3.3
Color States of Machine 2 During the Experiment
35
Figure 3.4
Total Number of Passes of a Trial
36
Figure 3.5
Total Number of Passes Summed of Sample Number
37
Figure 3.6
Plot of Number of Hits per Trial and Number of
Passes per Trial
39
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Page
Figure 3.7
State Transitions From Subject Choice to Future
Machine State
40
Figure 3.8
Distribution of Yellows for Machine 2
42
Figure 3.9
Distribution of Greens for Machine 2
42
Figure 4.0
Distribution of Blues for Machine 2
43
Figure 4.1
Distribution of Reds for Machine 2
43
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Overview
An apparent phenomenon which defies the theory of probability occurs
when Subject 2 plays this experimental game. He significantly exceeds his
probability of success, .25, by scoring over .29. The question that this
report addresses is: Is there a statistical or logical reason why he did
so well? The methodology used to attack this problem and the resulting
conclusions are summarized below. This summary can also serve as an outline
to this detailed report.
I.
Statistical Analysis of the Machine Experimental Data
Pre-experiment data analysis discovered a non-random characteristic
through the examination of forward-backward state transitions (i.e.,
Red-Blue, Blue-Red). However, the coefficient of correlation between
the forward and backward states of .58 for the experimental data, .49
for Machine 1 data and .48 for Machine 2 data were considered low
enough that this approach was dropped. Pre-experiment state transitions
had a coefficient of correlation of .93.
The experimental data randomness analysis consisted of examining
the distribution of color totals and the distribution of each color
taken over various combinations and permutations of the data. No evidence
of non-randomness was discovered.
II. Analysis of the Subjects' Data Responses
The subject's responses were analyzed with the emphasis on the
discovery of his strategy or the unveiling of a trend which would give
him a statistical advantage. The possibilities investigated produces
no solid reason how he was able to be so successful. However, in one
case there is a strong indication why he was able to succeed. It appears
that he was learning the states of Machine 2. The details of this are in
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
the remainder of the report.
Miscellaneous
The report contains a section entitled "Miscellaneous" for the
purpose of displaying detailed data which wasn't directly required by
the above more general analysis. Details such as how many successful
choices in the color red during the 50th trial were there, or what was
the relationship of the number of passes to the number of successes.
The terminology used is as follows: the term "trial" refers to the
string of machine states and corresponding choices from the time the
subject begins until he makes 25. non-passing choices. A sample is a
machine state and/or subject choice (including passes). There are
(25 + # passes/trial) samples in each trial.
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
I.
Statistical Analysis of the Machine Experimental Data
Forward-backward State Transition Analysis
SG11
In a previous memorandum (Memo ORD 2240-75, 12 June 1975 to
the question of randomness with the emphasis on state transitions as an indication
of non-randomness was addressed. The data used in the investigation consisted
of pre-experiment trials. The purpose of this section is to do a similar
investigation using the actual data which occurred during S2's experiment.
Table 1 presents all possible transition frequencies. All transitions
should have equal probability.
YELLOW
GREEN
BLUE
RED
YELLOW
204
199
199
216
GREEN
192
218
222
207
BLUE
211
206
228
222
RED
209
206
223
221
Restructuring into a two-by-six table as in Ref 1 produces:
Y/G
Y/B
Y/R
G/B
G/R
B/R
FORWARD
199
199
216
222
207
222
BACKWARD
192
211
209
206
206
223
The conclusion based on pre-experimental data was that these state-pairs
show a very strong relationship between forward and backward transition
:
frequencies (coefficient of correlation =. 93). However, computing the coefficient
of correlation,
Ps2
actual data = . 58,
it becomes apparent that the
degree of dependence is slightly reduced.
Therefore the dependence of
forward to backward states can no longer be considered as a strong indicator
of non-randomness.
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
The data used in the above discussion consisted of trials from both
machine 1 and machine 2. Since non-randomness, made apparent by the state
transitions, clearly existed for pre-experimental data, the investigation of
the experimental data continued to include a search for this trend in the
individual machines. The transitions (including identity) are as follows:
Machine 1
YELLOW
GREEN
BLUE
RED
YELLOW
96
79
88
92
GREEN
85
87
86
88
BLUE
85
82
90
87
RED
91
91
83
92
Machine 2
YELLOW
GREEN
BLUE
RED
YELLOW
108
120
111
124
GREEN
107
131
136
119
BLUE
126
124
138
135
RED
118
115
140
129
Computing the two coefficients of correlation,
=
.4934
machine 1
s2 data
and
= . 4838
a
machine 2
s2 data
it is obvious that the forward and backward transitions are even less dependent
than in the combined case. Thus ended the search for non-randomness through
state transition.
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
2.

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
As a by-product the following table is produced for general information.
BOTH MACHINES
MACHINE 1
MEAN
SD
MEAN
SD
MACHINE 2
MEAN
SD
FORWARD
210.8
10.7
86.6
4.27
124
9.74
BACKWARD
207.8
9.00
86.2
3.92
121
11.25
TOTAL DATA POINTS
3483
1446
2037
COEFF OF COV
.5843
.4934
.4838
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
3

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
3191.
2702
3650
Experimental Data Randomness Analysis
The machine data used during the S2 experiment has been combined, summarized
and/or permuted in an attempt to establish evidence or randomness or non-
randomness. If an obvious indication of non-randomness would have evolved
this task would be simplified because it would have become a closed form
frased
problem (i.e., the solution would be - the data has non-random characteristics).
However, what has resulted is that various forms of the data have been examined
with all indicating that the data is random.
Tables, plots and commentary are presented in this section to demonstrate
randomness and in some cases just to provide general information concerning
the machines data.
The distribution of the colors collectively and for each machine is
as follows:
Yellow
Green
Blue
Red
Total
Mean
Machine 1
365
353
356
372
1446
361.5
Machine 2
475
505
538
519
2037
509.25
TOTAL
840
858
891
891
3483
870.75
Machine 1 was not used in as many trials as machine 2 (44 trials to 56
for machine 2), thus the difference in totals. The standard deviation of
binomial distribution with n=3483 and p=1/4 is 25.56 which would imply that
each separate number is reasonably close to the mean.
Accepting the distribution of the totals consider the distribution of the
colors throughout the experiment. The popluations used for this investigation
consisted of the first 25 samples of each trial (100 trials total). This
population is acceptable since the distribution of its totals was reasonable
and since the performance of $2 was approximately the same (success-29.61%)
for this subset.
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
The following three approaches comprise the strategy used to attack
the question of color distribution.
1.
Each trial (abbreviated to 25 samples) as analyzed separate interval.
Obviously this will indicate any bias within each trial.
2.
The data (2500 samples) is divided into intervals of five samples
each.
This will indicate unusual repetitions either within the
interval or interval-by-interval.
3.
The data is reformatted into 25 intervals of 100 samples, where the
nth interval consists of the nth sample in each trial.
The results of approach 1 is shown in Figures 1.1.a, 1.1.b, 1.1.c, and
1.1.d.
The binomial distribution for this strategy (n=25 p=1/4) is mean 6.25
and the variance 4.69.
The plots indicate randomness throughout the 100
trials.
,
The results of approach 2 are similar to approach 1 and are shown in
the four tables in Figure 1.2. The plots indicated randomness but are not
. shown because of monotomy. The binomial distribution mean is 1.25 and the
variance .94.
1
The binomial distribution mean and variance for approach 3 is 25 and 18.75
respectively (Figure 1.3). A plot of the data (Figure 1.4) for the "RED"
case because of the concern for the higher variance and ranges. The 13th sample
seems to have an unusually high frequency of "RED" (44%). However in general
this investigation has not produced a significant non-random characteristic.
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
5

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
sample size
100
maximum
12
minimum
range
3
9
mean
variance
4.239494949
standard deviation
2.059003387
mean deviation
1.6314
6
median
mode
6
15M
M
M
M
M
()
0
0 0
1 OM
M
()
0.0
()
M. 0
0 0
()
.
00
0)
()
0 0 0 0
MO
0
MO 0 0 000 0
:0
0 00
0
0 00 0
0)
0
000 0
()
0 0
00 000
5M0 0 0 0
()
0
00 0
0000 0
()
0 0000
()
0)
0
0
M
()
0
0)
()
M
M
OM
M
M
20
M
M
M
M
M
M
80
M
M
100
40
6.0
1
Trial Number
Figure 1.1.a Distribution of Machine Yellows Over Trials
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
6
Number
of
Yellow
per
Trial
0 0 00
.
M
()
0 00
()
0 0
C
C
0
0
6.23

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
sample size
100
maximum
12
minimum
0
range
12
mean
6.13
variance
5.851616162
standard deviation
mean deviation
median.
2.419011402
1.9404
6
mode
5 7
15M
M.
M
M
0
M
0)
0
10M
MO
00:00
00
00
00)
M O
()
00
() ()
0)
()
00
M 0 0
00 0
()
0)
00
0)
0.000 0 0 0
MO 0 0
00
00
000
00
5M 0 0 0 0000 0 0
0
0
0 0
00 0
()
()
()
0 0.
0)
M
0 0
00
C
M
()
OM
M
M
20
0)
MOM
40
M
M
M
M
M
M
0
80
100
Trial Number
Number
of
Green
per
Trial
0 00.00
MO
0)
M
()
0
60
()
0 0
0)
Figure 1.1.b Distribution of Machine Greens Over Trials
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
7

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
sample size
1 00
maximum
11
minimum
1
range
10
mean
6.21
variance
5.218080808
standard deviation
mean deviation
median
mode
2.284311889
1.8194
6
5
15M
M
M
M
M 0 0
Number
1 OM 0
()
.0
0 00
0)
()
of
Blue
M
()
0
0
()
0 00
00 0
()
Per
Trial
M 0)
000
0000
0
C
()
000
0
0
0000
0)
0 0 0
0
M
OM
M
M
20
M
M
40
M
60
M
80
M
M
0
100
()
0 0
()
M
0 0 00
()
000
()
0
()
000
0)
00
()
5M
MO
M
MO
00
C
0
()
:
()
0)
M
M
()
Trial Number
ApproveNingarRelease2003041ButGIARDP96-20787000209150011-4
8
M
()
0
00
0 0 000000

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
sample size
100
maximum
12
minimum
1
range
11
mean
6.43
variance
4.631414141
standard deviation
2.152072058
mean deviation
median
1.7158
6
mode
6
15M
M
M
()
M
0
MO)
1 OM
00
0)
()
M
00 00
000
00
(
00 0 00
MO)
0
0000
0 0 00
0 0 0 0 00 00
00 00 0 0
M
5M
00
00 0
()
0)
0
()
0)
MO
()
()
0 00
()
0)
0)
()
M
M
20
M
M
40
M
M
60
M
M
80
M
M
100
0
Trial Number
Approved Fjor Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Distribution of Machine Reds Over Trials
9
Number
of
Red
Per
Trial
0
()
0 0 0
0 000 0
M.
0000
()
00 0
0 0 0
0 0
0)
M
M
OM
0)

Approved storiRelease 2003/04/ 08 : CIA-RDP96-00787R000200150011-4
sample size
500
5
maximum
minimum
0
range
mean
variance
standard deviation
mean deviation
median
mode
5
1.246
0.9594028056
0.9794910952
0.784848
1
1
Distribution of Green
500
sample size
5
maximum
0
minimum
5
range
1.226
mean
0.9969178357
variance
standard deviation
0.9984577285
mean deviation
1
mode
1
Distribution of Blue
dstat grp ;< 3:
sample size
4
minimum
4
range
1.242
mean
0.95/3507014
variance
standard deviation
. 0.9784429985
0.792/92
mean deviation
median
mode
Distribution of Red
sample size
500
maximum
5
minimum
0
range
5
mean
1.286
variance
1.026256513
standard deviation
1.013043194
mean deviation
0.823216
median
mode
1
1
0.804512
median
500
maximum
U
Figure 1.Appisted MoriRelease03/8HARD Time
10

Approved For Release 2003/04 98
Yellow Distrib
CIA-RDP96-00787R000200150011-4
maximum
31
minimum
19
range
12
24.92
mean
variance
10.57666667
standard deviation
3.252178757
2.6304
mean deviation
median
mode
24
24
Green Distribution
sample size
25
maximum
35
minimum
15
range
20
mean
24.52
variance
24.59333333
standard deviation
4.959166597
mean deviation
3.9392
median
25
mode
22 25
Blue Distribution
sample size
25
maximum
34
minimum
19
range
15
mean
24.84
variance
14.47333333
standard deviation
3.804383437
mean deviation
2.9664
median
25
mode
26
Red Distribution
sample size
25
maximum
44
minimum
16
range
28
mean
25.72
variance
26.71
standard deviation
5.168171824
mean deviation
3.3664
median
25
mode
25
Figure 1.3 Distribution of Machine Colors When Samples are Taken 100 at a Time
(One From Each Trial)
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
n

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
45M
0)
M
M
M
40M
M
M
M
M
35M
M
M
M
M
()
30M
()
M
0) 0
M
Number
M O
00
of
Reds
25MO
0)
00
.0)
M
()
()
M
0
M
()
0)
()
M
20M
0)
()
M
M
M
M
()
1.5M
M
M
M
M
M
M
0
10
20
30
Sample Number
Figure 1.4 Distribution of Machine "Reds" when the Samples are taken 100 at
a time (one from each trial)
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
12
M
()
0
M

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Approach 1 has been repeated for Machine 1 and Machine 2 separately to
check for abnormalities. The binomial distribution mean and variance are
as follows :
Trials
Mean
Variance
Machine 1
44
11
8.25
Machine 2
56
14
10.5
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
13

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Machine 1
Machine 2
'Yellow
sample size
25
maximum
19
7
minimum
7
9
range
12
11.4
mean
variance
7.51
2.783882181
standard deviation
2.740437921
2.224
mean deviation
2.176
14
median
mode
15
sample size
25
sample size
25
maximum
17
maximum
24
minimum
4
minimum
(
range
13
range
16
mean
variance
10.68
mean
12.72333333
standard deviation
3.118760438
standard deviation
3.56697818
mean deviation
2.3584
mean deviation
2.7808
median
1.1
median
13
mode
1
mode
13
sample size
25
sample size
25
maximum
15
maximum
25
minimum
3
minimum
10
range
12
range
15
mean
variance
7.726666667
variance
8.943333333
standard deviation
2.779688232
standard deviation
2.990540642
mean deviation
2.3072
mean deviation
1.984
median
mode
8.12
mode
15
sample size
25
Red
sample size
25
maximum
19
maximum
21
minimum
4
minimum
11
range
mean
15
range
mean
10
variance
10.5
variance
10.01
mean deviation
2.4
mean deviation
3.163858404
2.6624
median
1.2
median
13
mode
12
mode
11
13
standard deviation
3.240370349
standard deviation
14.52
10.32
mean
14,12
11
median
14
12
1.2
Green
8
13.84
9.726666667
variance
13.52
7.75
sample size
maximum
minimum
range
mean
variance
standard deviation
mean deviation
median
mode
25
16
Figure 1. ApprevedeF06 Release, 2003/04/18 : CIA-RDP96-00787R000200150011-4
tribution for Machine 1 and Machine 2 on a Trial
To Tr
Trial Basic
(14) ..
11.6
Blue

1
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Best Strategy
Based on the above analysis what is the best strategy to pursue? No
good strategy is available based on the randomness of the data. The best
possible strategy based on the above transition matrices is:
1.
If the subject can't distinguish between machine then press blue
when blue appears, else pass.
2.
If the subject can distinguish them on Machine 1, press yellow
when yellow occurs, and on Machine 2 press blue when red occurs.
For all its worth, of the existing data the following success would
result - 26%, 26%, and 27%.
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
15

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Analysis of S2 Data Responses
The attempt here is to discover a reason for S2's success at responding.
The investigation was unable to give a definitive reason for his success.
Although no strategies were uncovered there was in one case a indication that
the subject was learning.
Two major approaches have been taken in this investigation.
They are .
as follows :
Ther detine
1.
Strategy of S2 - Was there any trends in the way he guessed? Did,
he respond based on the previous state of the machine?
2.
Hit analysis - Did the subjects' hits (correct choices) increase
within a run; did it increase from run to run (i.e., was he
learning?)
Strategy of S2
For general information and future reference the first figure (Figure 2.1)
presented is the actual choices. One item of curiosity from this is that
when he passes, he tends to do it in strings. This characteristic of course
wasn't pursued because of its insignificance to this report; however, observations
like that are pointed out throughout the report as possible importance to those
in the field.
Total Color Choices
The distribution of $2's color choice totals are shown below.
Approved For Release 2003/04/18 : C6A-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
0210232010213003020300330
0203121303030330000102332
3003103030312032103222123
0233310020320130300020313
3030030010303031313030103
3303031303030003202103103
0323030303020301032030330
0320303030302103030301303
0303032022303010313021020
3010103103013303013023013
0313023313303102013103231
0210310310310332031030230
3030203103030130130303023
3030323013030203010330303
3030030302303130313031300
3023130302102313010130203
303070307300103077230770731377030773
320301303077307070130303723770373
03023010737037737301730307177707207370
021303077730702302303070723730703
03701037777321033700371307077301031
0777377730777317077377037233103273073030373
373031377773073277307707307707073007077773203
307370307302130313313777073023777377770
03031700120120313027772323103
0131320203120310773071730777772031
30373030377730301307307770330377777773070
31217033030130037777771300012003
00273077037727777731077737777773777777773132133013070
31737777777777777170710777777730137777073703132777777030777737013
377770777701777777770307373177777303031031031020
3771727077770130717371777737777777777702030317013201
0377777777773770103777777077707777777773013131303230320
00230713013077777777777713013023201303
077777701010203010230703730270730777777713
307137770370777037777732310777777777777777031307777777777777773703703771
302077771303703130313021037013777777777777770
31037321013013102310370107731
31313023130132013023730177703
13037373730130132077777377777777707313021071
137310373731730217727317713177777777703733170
13237013077072313103127773713173777373
31377777777777377703313102137177777717077731727120713
01237073773177731737201720307072170130
073373113701310701077201377032770070.
321317032331303203723032123
137370710303107720311307100323773
10307710237371307307230233203730
2030330231313302212121331
23077701273212000303333130300
Figure 2.1 Subject 2 Color Choices for First Fifty Trials (0-yellow, 1-green,
2-blue, 3-red, 7-pass)
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
17

323031301202137077123217
Release 200
2003/04/18 : CIA-RDP96-00787R000200150011-4
031137389773162037777777770377373017333070
20703123070231703030330133703
30170102031730730300330313713
01007777373707303777173777273377310770777130777373773
20707 370731 777370077037370002731 777707077777073777777377773
01730330320370330327013703013
303717033207303073773013023737203
0373737303373032173233377173733707371
23707377313033333703773773707173777377373270
1310730373773010333337073731 3707700
3337370773073037 33331 3037 3370707770733
033037737703337337777077777327777301027337333
0303037200003773733777077377 33030332
33032133270323233130121330
10701101301101313030230123
300703300723730030371777137777033002
031 3373737030003200030001003
302 3321 31000001371303703037770
772017710077030723703137310137773771 77777303
337171017711371300217333733030733
30717373717077130117303707301 373370071
03303203020071027107377121270703
03231327320373023770331110077700
33173707371071317331331730117207073
271313107327033277731177130323303
300037 3300033003710303071 330
30127001 333301 30777 37077373303377770770
0303703037073732311370710732001773
377 33070072000770300373130003002
1 32002000300303770300731723370
30707207020773307033030303777377737377073
07077730703700377777707731707330307307077770737373
00707737377003073730777777707377777737773770300077773333
1 301037132010717301002720073723
3101310317001300001730073020
03777720070773100770707373007200730700700
30300007100000232113002002
3031 301 301 3201 30231033003
2301203130120310311303120
30130231031737130730323001 31
301 301 301 3201302101302303
130231032303713273031030130
3010310310773230313073021 331
310313031737701373001 330033777713
31301030310330307377070037717003
02313033201 37001 37230201 330
0217373103101303700073027777310373
1 37073107103702373132710331073703
331300301707301070700371073700713
Figure 2.1 (Continued) S2 Color Choices for Last 50 Trials
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
18

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Yellow
Green
Blue
Red
Total Times Chosen
881
411
237
971
% of Total
35%
16.5%
9.5%
39%
The first inclination is to try and determine how his strategy of choosing
so many yellows and reds benefitted him.
Examine the following table:
Yellow
Green
Blue
Red
Total Number of Hits
255
127
60
50
292
% of Total Hits
35%
17%
8%
40%
% of Success in Color
29%
(Hits - Correct Choices)
31%
25%
30%
As can be seen his results with blue are significantly lower than the others.
However, assuming the probability of success to be .25 and using the binomial
distribution the expected value =69 and the standard deviation = 7. The
inference from this is that the 60 Blue hits are not a statistical abnormality.
However, it is curious that he did so much worse on his lowest preference.
State Transition Color Choice
This investigation consists of examining the states of the machine verses
the choice on the next sample of the subject (i.e., if the machine shows "red"
does the subject consistently choose one color on the next turn). Consider
the following table:
SUBS
Machine
MACH
Yellow
Green
Blue
Red
Pass
% Pass
S
Yellow
106
119
69
314
210
26%
U
B
Green
177
2,5
69
316
252
30%
J
E
Blue
241
99
27
198
302
35%
C
T
Red
322
157
65
97
218
25%
V: , 30
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
19

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
The subject obviously avoids repeats (i.e., he assumes the machine
Won't repeat a color) which, based on the machine data analysis, isn't a
strategy which would give him a statistical advantage. Previous analysis
showed that identity transitions are approximately equally probable as
nonidentity. Notice also that he passes 35% of the time after seeing a blue.
The same state transitions are shown below separated by machine.
Yellow
Green
Blue
Red
Pass
M
Yellow
48
49
25
150
83
A
Green
62
13
35
153
83
H
I
Blue
105
36
10
78
115
E
Red
133
72
30
58
64
1
W2.94
M
A
C
Yellow
58
70
44
164
127
Green
115
12
34
163
169
I.
N
Blue
136
63
17
120
187
E
2
Red
189
85
35
39
154
The negative state transition (i.e., relationship of the subject color
choice to the machine state on the next sample) is considered too bizarre of
a concept to be presented in this section. Results of that investigation
is found in the section entitled "miscellaneous"
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
20
C
N

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Hit Analysis
This section is significantly more important than the randomization
analysis of the machine data. The reason is that if he is not learning from
the machine or he is not taking advantage of biases then the discovery of
such non-randomness is of little value to the overall analysis.
Learning from Trial to Trial
The question of whether the subject learned from trial to trial can
best be answered by examining the following three plots .. The first is the
number of hits vs. the trial number, the second is a frequency distribution
of the number of trials vs. number of hits, the third is the accumulated
probability vs. the trial number.
Approved For Release 2003/04/1821CIA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
15M
M
EE
EE
0
0
0
1 0M
M
00
I
00
M. O
0
0
000000
000
0
00
MO
0
00
00
0
0
0 .000
M.
0
0
000 000
0 00
5MO 0
0
0
M
C
M
M
0
M
OM
M
M
20
M
M
M: .
M
E
M
M
M
0
40
60
80
100
Trial Number
Figure 2.2 Plot of number of hits/trial
25M
M
M
M
M
20M
E
M
M
M
15M
M
M
of
Hits
M
per
Trial
M
M
M
M
5M
M
M
0
C
M
M
0
OM O M
M
M
M
MO 0 0 0
OMO
0 0 0 0MJ 0 0
0
5
10
15
20
25
Figure 2.3 Frequency plot of Number of Hits
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
22
Number
of
Hits
Frequency
10M

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
0.30M
M
0.0
()
0 0 00
()
000
0.29M
0 000
()
000
()
0 00 00
() 0 00
()
()
0 000000 0000
C
0000000 00
()
0000
0)
0 0)
0
()
M
()
M
M
0.28MO 0
()
0)
()
0.27M
0.26M 0)
M
M
M
0.25M
M
0)
M
M
M
0.24M00
M
M
M
40
M
M
60
M
M
80
M
M
0
100
:
Trial Number
Figure 2.4 Cumulative Success Ratio of Subject (both machines used)
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
23
Cumulative
Probability
of
Success
M
M
M
0 0 0
M
M
M
M
M
M
M
M
M
M
()
M
20

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
M
M
M
M
0.31M
M
M
M
M
0
0.30M
M
M
M
()
()
M
0.29M
()
M
()
()
00
Probability
M
.. 00
00
00
00000
M
()
M
0 0
0.28MO
00
M
0
M
M
M
0.27M
M
M
M
M
0.26M
M
M
M
M
0.25M
M
M
M
M
0.24M
OOM
M
M
M
M
M
M
M
M
M
0
10
20
30
40
50
Trial Number
Figure 2.5 Accumulative Probability of Success on Machine 1
1
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
24
Accumulative
00
1

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
0.30M
M
M
Q000
M
00000
00
M
00
0.29M
,00
()
M
M
00
M
M
0.28MQ
M
M
()
M
0)
M
0.27M
M
M
M
M
0.26M
M
Accumulative
M
M
M
0.25M
M
M
M
M
V
V
0.24M 00 M
M
M
M
M
M
M
M
ME
0
10
20
30
40
M
50
M
60
Trial Number
Note: V - Points at which he switches machines
Figure 2.6 Accumulative Probability of Success on Machine 2
>
L
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
25

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
The first plot (Figure 2.2) demonstrates the randomness of the number
of hits while the second plot (Figure 2.3) demonstrates the frequency
distribution takes on a "normal" appearance. The accumulative probability plots, at
first glance, indicates that the subject was in a learning mode for the
first five trials. A closer examination of the data indicates that this can
occur naturally as part of the statistical distribution.
The first three number of hits points are 7, 5, and 6 considering the
first 75 points as the population with probability of success = . 2936
(the final probability) then the expected value is 22 (using binomial
distribution) and the variance is 15.55 (S.D=3.9). As a normal deviation
from the mean (i.e., using normal distribution approximation P(x<18) =. 13.
Although the observed learning can be rationalized as a natural statistical
deviation it warranted further investigation. The plots of the accumulative
probability of success for machine 1 and machine 2 are presented in Figure 2.5
and Figure 2.6. The plot for machine 1 (Figure 2,5) is a typical sinesodial
decreasing amplitude convergent curve. The plot for machine 2 however, is
very suspicious in terms of learning. The major peaks of the curve (at
approximately trial 10, 23, 40 and 56) are increasing which implies his
probability of success is continuing to increase instead of converging on
one point. Another interesting pointsis that the points at which he switches
onto machine 2 are 1, 9, and 36.
Also of concern is the sharp upward turn during the last 8 samples.
The hits totals for this period, starting at sample 49 is 10, 10, 8 11, 6,
8, 7, and 11 for a total of 71 hits out of a possible 200 for a probability
of success of .36. Once again using the binomial distribution and using the
probability of success of .29 (the
cumulative probability up to the 49th
point) the expected mean is 58 and the standard deviation 6.42. Using the
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
26

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
12.5M
M
0
()
M
M
M
M
()
M
M
M
0 0 0
()
0
C
0
00 0
00
0
0)
()
7.5M
MO)
00.
()
00 0
0 0 0
()
M
M : 0)
M
()
0)
()
0000
()
5.0M 0 0
()
()
()
M
M
M
M
2.5M
M
()
M
M
M
0.OM
M
M
10
M
M
M
M
M
M
M
M
50
60
0
20
30
40
Trial
Figure 2.7 Plot of Number of Hits on Machine 1
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
27
Number of Hits
()
C
10.OM
()
00
0.
00
()
M
M

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
normal approximation the probability P(X 71) =. 02 of such an occurrence is
quite low.
Although there are only 56 data points in this population and the
apparent abnormalities are statistically possible (with low probability) this
investigation concludes that the subject's learning for this case must be
flagged as a real possibility. Figure 2.7 (Number of hits on Machine 1) has
been added to provide clarity. It appears that the subject just didn't have
"low hit" days toward the end.
Learning within a Trial
The question of learning within a trial or run has been investigated
bath machines?
by summing the number of hits of the Ith sample for the run. The results
are somewhat distorted because of the inequitable distribution of passes.
The lower numbered samples have significantly more hits because of this.
2,5?
A plot of the number of hits per sample vs. sample number is shown in Figure 2.7.
Notice that the first sample has a value of 34 hits. This means that
everytime he ists down for a new 25 sample trial he hits 34% of the time on
his first try. With this in mind along with the rest of the data points, it
is obvious that the subject doesn't learn throughout the trial.
Approved For Release 2003/04/18 : CiA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
40M
M
M
MO
M
30M
()
()
M.
M
C
0 0
MOO
00
M 000
0 00
Number
of
Hits
20M
()
M
00
0 00.0
C
M
()
M
0 0
M
1 OM
000
()
0
M
0 00
M
00 00000
M
M
M
M
MO 0000000000000
0
M
20
40
60
80
Sample Number
Figure 2.8 Total Number of Hits Within a Trial
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
29
M
()
00 00
OM

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Miscellaneous
Numerous arrays of data have been examined for the purpose of obtaining
some insight into the data.
Some of the data is being printed herein so that
the data can be examined more closely if desired.
This first table is presented for use as a quick reference.
Day
Last
Trial
Number
of
Tracks
Machine
Used
1
8
8
2
2
16
8
1
3
24
8
2
4
36
12
2
5
44
8
2
6
52
8
1
7
56
4
1
8
64
8
1
9
68
4
1
10
72
4
1
11
76
4
1
12
80
4
1
13
84
4
2
14
88
4
2
15
100
12
2
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
30

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
The following displays are presented below with little commentary.
I.
General trial summary (Figure 3.1). Each trial (25 choices) is
listed with the following information.
A.
Machine used (1 or 2)
B.
Total number of machine states in each color (i.e., 6 yellow,
6 green ..
.) for each trial.
C.
Total number of subject choices for each color for each trial.
D.
Total number of hits for each trial.
E.
Total number of passes for each trial.
F.
Breakdown of hits by color.
II.
Machine data for machine 1 and machine 2 separately (Figures 3.2, 3.3)
Just by examining these displays it may be possible to glean
meaningful information. For example, machine 1 was used for the
first 8 trials during which the first state of each trial was a
yellow or red. If the first sample of each trial is most memorable,
perhaps this is responsible for the subject's obvious preference of
yellow and red (see Section 2 - Analysis of S2 Data Responses).
7
III. Plots of the number of passes made.
A.
Number of passes vs. trial number (i.e., trial is 25 or more
samples) (Figure 3.4)
B.
Number of passes vs. sample number (Figure 3.5)
31
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
yel arn blu red hits pas yel grn blu red
mach mach mach mach sub sub sub sub numb num hit hit hit hit
trial mach yell gren blue red
1
2
0
00
24
11
-
3
5
6
7
0
3
0
4
2
2
5
2
1
8
7
2
2
5
6
4
10
4
7
10
4
3
4
8
8
0
4
1
1
2
1
7
2
m
0
5
7
7
6
5
4
3
a
10
3
2
10
5
0
0
0
0
3
2
4
3
0
3
0
0
0
2
12
V
1
7
2
7
9
9
10
10
2
2.
-
0
0
1
2
0
4
3
3
5
19
N
8
9
10
-
11
3
a
0
00
-
-
8
13
8
1.0
18
20
14
+
a
16
4
1
3
3
w
0
2
3
1
4
6
2
1
1
0
1
2
20
21
CV
7
13
5
8
-
1
4
9
0
13
7
0
5
7
8
2
a
7
28
40'
23
27
30
13
17
3
2
N
N
2
3
1
0
1
0
1
0
1
5
32
2
12
19
C
20
10
a
5
3
8
3
6
7
7
47
20
2
0
4
3
3
4
1
-
0
0
O
2
42
N
11
-
7
0
7
3
11
43
~
15
04
11
a
8
45
-
12
00
a
7
4
+
6
11
0
N
0
2
4
47
1
a
0
0
4
8
6
2
9
8
3
2
0
1
48
-
9
10
7
6
8
3
4
10
7
2
1
1
2
49
1
7
10
0
2
4
0
6
9
7
0
0
5
1
1
50
-
9
12
-
7
9
3
4
9
0
4
3
0
0
3
7
10
m
4
8
5
0
2
1
0
0
2
6
8
0
0
2
2
1
0
-
-
-
2
-
-
8
a
3
2
U
2
0
1
10
10
O
a
a
1
4
10
1
5
6
1
5
7
5
4
11
Inno
10
8
12
10
3
5
2
2
1
4
1
2
2
8
10
11
8
a
5
7
CO
0
3
2
1
6
3
16
17
2
10
1.2
9
7
a
7
11
a
-
0
22
2
13
2
NOO
9
8
-
24
2
10
a
11
25
N
w
-
to
15
a
=
1
28
2
5
0
10
-
14
8
4
3
0
30
N
0
19
18
60
12
8
0
6
5
7
+
1
1
3
2
10
0
7
10
8
9
NON
5
2
0
1
1
4
2
4
37
2
5
16
13
(
-
9
மி
2
9
6
8
10
40
N
10
14
0
12
6
7
4
4
6
6
4
2
7
0
5
-
8
13
28
13
-
2
0
4
5
0
1
3
4
0
4
1
1
2
1
44
N
10
-
+
-
4
9
3
9
7
a
a
4
4
+
19
20
1
1
2
1
0
0
2
3
4
3
39
222
5
7
-
7
890
0
10
8
6
6
8
2
3
2
2
10
11
41
N
+
17
19
22
9
4
1
11
12
1
1
1
5
3
29
N
N
16
16
1
4
31
2
10
16
a
a+++
10
8
15
12
4
2
5
4
0
-
12
7
7
13
26
2
10
3
27
2
11
a
-
a
12
a
a
10
8
8
8
10
5
2
8
1 00
33
N
11
34
2
35
NV
w
2
2
4
4
4
3
0
23
2
a
a
0
N
7
00
9
8
7
8
10
a
6
0
n
0
1
7
a
10
a
0
N
00
0
1
1
0
0
0
O
0
2
w
A
1
4
15
1
0
5
2
0
0
0
0
1
18
N
4
N
a
9
2
0
0
2
0
0
1
2
0
2
19
12
00
5
2
1
3
5
6
1
2
2
8
10
0
12
-
7
0
2
2
8
8
11
1
0
3
3
in
3
2
10
0
4
-
11
3
7
3
2
8
7
4
0
7
8
10
46
1
5
11
w.
11
- a
7
.
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Figure 3.1 Selected Parameter Totals Listed by Trial Number
32
1
:
1
1
0
1
1
0
7
10
-
a
36
0
38
4
7
00
-
1
3
4
5
6
2
6
4
4
3
3
5
4
3
2
10
0

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
mach mach mach mach sub sub sub sub numb num hit hit hit hit
trial mach yell gren blue red yel grn blu red hits pas yel grn blu red
51
1
0
n
10
8
0
5
0
CO
+
2
2
3
1
7
15
11
9
89
m
3
10
V
3
1
-
1
1
-
5
7
6
12
9
5
1
10
1
12
2
2
-
3
3
5
2
8
4
0
12
16
16
13
17
000
8
8
8
7
8.
1
2
3
0
0
1
0
0
63
1
10
18
0
7
6
1
2
16.
4
8
11
1
4
1
3
0
0
1
4
0
0
-
C
2
4
5
1
67
1
8
10
10
5
a
13
2
1
a
-
3
4
0
0
0
4
70
1
9
6
12
7
8
5
8
7
0
10
m
CO
13
7
2
1
1
1
0
0
4
-
2
73
1
11
0
in
10
10
8
4
4
a
or
10
1
3
4
0
V
78
1
9
9
10
-
a
3
1
12
10
8
8
7
CC
10
8
9
7
5
16
25
31
6
3
2
A
7
2
3
1
-
83
2
7
10
17
16
13
-
0
-
ONON
m
0
n
1
2
88
.00
m
00
14
3
4
4
7
-
0
4
2
1
1
1
N
90
7
7
10
7
2
7
6
2
10
0
w
4
1
1
-
93
6
7
7
7
7
5
3
C
10
2
3
2
1
4
94
5
6
+
13
7
6
V
10
10
3
3
1
1
1
0
5
96
2
7
9
7
9
11
4
0
10
1
0
7
5
1
0
n
97
2
8
8
0
5
8
4
4
9
2
2
1
1
2
98
7
12
10
5
9
5
2
0
9
3
3
0
2.
99
8
a
8
8
7
0
2
0
87
8
2
V
0
4
100
a
5
9
10
U
5
0
3
11
8
5
3
0
3
1
1
1
1
1
1
1
10
8
9
0
0
7
1
0
4
4
9
2
2
5
2
2
13
12
10
10
00 00
8
1
11
3
1
0
4
m
69
1
10
8
4
8
10
4
2
a
a
5
4
0
1
2
0
2
71
1
11
N
a
13
9
4
5
0
a
7
4
0
1
2
1
3.
75
9
-
-
7
8
5
8
4
6
11
10
0
00
2
7
0
2
0
3
3
3
80
1
8
6
10
8
14
1
12
2
3
82
2
6
14
10
-
11
0
2
12
84
14
12
16
-
12
va
0
C
6
6
87
13
4
13
a
84
7
7
4
7
0
0
1
3
1
-0
92
4
6
10
5
8
6
3
8
6
0
1
3
1
5
95
7
5
0
11
7
6
C
12
00
8
2
0
0
C
5
4
1
85
7
7
10
7
0000
17
1
2
5
8
89
N
91
NNNNN
11
7
4
0
12
2
-
0
79
1
7
8
7
N
a
4
2
2
81
V
13
4
8
17
8
6
2
1
9
12
4
7
19
3
0
1
10
11
0
9
10
65
1
7
a
2
8
66
1
m
12
+
7
8
O
1.1
5
-
11
-
3
2
0
2
3
1
0
+
14
12
14
8
14
12
2
a
3
1
4
58
6
-
11
4
-
N
8
2
1
1
1
1
0
4
4
4
7
4
62
1
13
12
14
2
19
11
-
8
.
15
11
-
0-
4
4
2
0
5
a
amparo
V
28
34
4
2
0
3
3
0
1
2
59
60
61
13
0
1.2
8
8
76
1
8
-
+
10.00
0
7
77
1
-
w
w
+
w
2
5
2
1
1
0
1
0
1
NNNN
NNN
Figure 318 : CIARE#96-00787R000200150011-4
Approved For Release 2003/0
33
1
10
4
72
1
-
74
1
+
11
1
4
0
0
4
1
-8
7
5
2
10
0
12
19
10
13
20
0
0
2
0
5
55
56
57
8
5
-
00
C
64
68
2
3
0
1
0
0
0
1
0
2
5
13
0
86
12
1
NO
9
5
5
0
C
7
2
2
52
53
54
7

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
0031003121303211033331132
0113100111033230023033300
2232103103310123321302022
2203311021022233333312000
3030220232231003033033122
3131202300133133132213001
2221122313233202200312210
1221131110300132213003120
03221123202221101003203120123102330031
001122313200131033002212023000130131
130223130102313222122130302
211120010202223111321220321100300
23333030020021221111031210110120
3221121001210212010100131
11111330101033101300310231010
01202123323202222312033013130
103112123203303111022312321102211012313210
13310102231200200000123003322
32133133233120023230021303203
21131110302330311003000003013331333122220122211132022
32220232333331022012010231302232200233201313111112112200012
30021020322232132003022203323
122323003321332212001303302233232
0222003302023212023123203010022221201
22121100300031311303030212303222031112021323
03221320000232112031322130301030211
10110332022100001033330021020223232303
033111022302000012112311223111211233110100121
020311131021101301033122001233320103
13133011310013013133020102
10103132133321232111110113
330201102012032131231212210032203113
1230200002033210310031203333
101130032201331300330301020121
30132032333032330033012022323213303123021212
112233022001200303031000123121033
13221230023032101331321221230112202203
10122031320033013313300100200233
33120221103320211133311121122013
31022132112201313120103131103032000
231011112232011111031000311201303
0000223302302002321000233112
302310321323331121020131200010232130223
3313011313201303310123210320033222
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Figure 3.2 Color states of machine 1 during the experiment
(0-yellow, 1 green, 2 blue, 3 red)

3030021330303023133311131
3323130A010Ved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
0303111222003203201112101
0220233110022220313200212
3230122322222001002111212
0323130003013313030203112
0121131322332323102110233
3000323233210023033221323
111310320122303013031012200113012201
022322300312323333132311110312211
01323231132020102230201233321331131020
321301110321331100202112310033111
00232122130302000121301333121323021
1101020113302031202031200203310013211132320
223222222001010001322110322323112331333301203
212131201330120310020213313002002222133
22023133332321021111311012311
2200211300320330002023133220232133
10023020000130321222213220132222220313033
33323012112033222122030123312303
13033112201002231233321122113312311121302202132013322
00212121201212003111312232223230002213311203101120113013032012300
220133203233331331032310103310300121032031233123
1233013220323210222202330110000232023231211202021323
0032132231010233122322332121111322102103202202120012120
30212022020230003000222003011002301323
330220332330223033222300033123113111213020
230031110213332123232010333012233111313322001230203301100322113213222302
331100231210122132011131113222211223203321331
31211032332213300312123232100
00203100132132212112233013202
12201200130132103221031100101233110223033321
210311120033110330112003302313112232211002313
13121022332310310100311100323000201213
20023011200113201121213130102210003223312203030133022
22130202313302320310213012300212
123000000103011020322022203203
11221232201333322321010213310311121013103
22122330301301101323312132023302212233221322132330
33330202202003321001220123121120131212303301203201213302
1303220101302331233021222201012
0323332110010001012102301030
11300023311100103001221112122201321302000
20020333331232021123023032
1102302323213122203102030
2103010003333212110310132
1011021230212001120111200203
0232212112223203231310120
033231031211110302023012232
1130230333333230211113233300
203213203111330213320330022222303
31303012320132312113320110032012
212203312300210001010131321
0320102302131222201103111110312212
111303223212300102300230301122113
r Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Color states of machine 2 during the experiment
(0 yellow, 1 green, 2 blue, 3 red)

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
50M
M
()
M
M
M
40M
0)
M
M
M
0)
M
30M
()
M
00)
()
()
()
M
()
M
20M
()
000
0 0 0
M
()
0)
()
M
C
M
0
()
00
()
0 0
M
()
()
0 0 0
()
M
00 0
()
0
:
()
00
M
0
00
000. 0
()
0 00
OM00000000 M
M
M
40
()
M
60
M
M
00
M
2
20
80
100
.
M
Number
of
Passes
C
C
1 OM
0 0
0
0 0
000 0
0 0
M
0 0 00
M
()
.
Trial Number
Figure 3.4 Total number of passes summed over a trial
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
0)

35M
M
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
M
()
0
C
M
()
30M
M
()
M
()
M
0 0
()
M
25M
0
00
()
M ()
00
000 0
0 00 0
M
M
M
M
M
M
()
()
0)
20M
()
()
.
M O
()
0)
15M
M
M
M
M
I OMO
0)
M
M
()
00
M
5M
0)
M
000
0 0
M
0 00
M
0)
00 000
MO
M
M
M
M
M
M 000000000
0
20
40
60
80
Sample Number
Figure 3.5 Total number of passes summed over sample number
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
37
Number
of
Passes
0)
M
0)
OM

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
C.
Number of passes and the number of hits vs. the trail number
on one plot. Investigation of the hits/passes relationship
was dropped when the coefficient of correlation between the
two was computed at -. 114
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
38

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
50M
M
()
M
M
M
40M
M
()
M
M
()
M
30M
()
M
00
0)
0)
M
M
M
M
M
C
20M
000
00 0
()
()
()
()
M
.0 0
()
00
()
0)
()
M
PO . P
P
-
PO
0 0 0
P P
1 OM
PPP
PPPP
PP
PP
()
P PPPOP
P O
MPP PP P PP P PP · PP
POPP POPPPPP POO PPPPP
OP()
MPPP
P P PP
P P
PPPP PPP
O P .P .O P PP
P
M
OOP
000 0
P OPP
()
()
00
M
()
P
P
()
()
()
0 0
OM 00000000 M
0
20
M
M
40
()
M
60
M
M
80
100
Trial Number
0 - passes per trial
P - hits per trial
Figure 3.6 Plot of number of hits per trial and number of passes
per trial
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
39
Pass
and
Hit
Total
P
00
M
.
0
()

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
IV.
Tables of state transitions which reflect the influence of the subject
on the machine. For color choices of the subject the table shows the number
of colors the machine has on the next sample. For example on the first table,
when the subject picked yellow, on the next sample 197 times the machine
state was yellow.
MACHINE STATES ON FOLLOWING SAMPLE
Yellow
Green
Blue
Red
Yellow
88
77
87
95
Green
38
46
39
47
Machine 1
Blue
27
28
24
24
Red
120
105
99
112
Pass
84
83
98
81
Yellow
109
124
128
141
Green
58
47
58
66
Machine 2
Blue
25
32
42
30
Red
121
125
136
102
Pass
146
162
161
168
Yellow
197
201
215
236
Green
.96
93
97
113
Both
Machines
Blue
.52
60
66
54
Red
241
230
235
214
Pass
230
245
259
249
Figure 3.7 State Transitions from Subject Choice to Future Machine State
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
40

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
V. Because of the possibility that the subject was learning the state of machine
2 the distribution of the colors are plotted in Figures 3.8, 3.9, 4.0, and
4.1.
The only states used are those in which the subject didn't pass.
Therefore there is a total of 25 for each trial.
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
41

Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
10.0M
0)
C
M
M
M
M 0
00 0 0 0
()
.0 0
7.5M
MO
000 0
M
MO
()
5. OMO 0 0
C
000 0. 0
M
M
M
0 0 0
0 0 0
M
M
()
0)
0
0
0
C
2.5M
M
M
M
M
0.OM
M
M
20
M
M
40
M
60
Trial
Figure 3.8 Distribution of Yellow for Machine 2
12.5M
M
M
M
M
10.0M
()
M
()
0 0
M
M
MO 0
00 0
00
00
5.0M
M
0 0 00 00
0)
()
M
M
0)
()
2.5M
()
0)
M
M
M
M
0.OM
M
M
20
M
M
M
0
40
60
Approved For Release 2003/04/18 :rCIA-RDP96-00787R000200150011-4
Figure 3.9 Distribution of Green for Machine 2
42
2
MO
M
MO)
0)
0 0
0)
00
7.5M
0 .0
00
()
M
0 0 00 0
0 00 0 0 0
Number
of
Green
0 0 0
0 0 00
()
Number
of
Yellow
M
0

12.5M
M
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
:
M
10.OM ()
0
C
M
M
M
M
()
00
()
C
0 0 00
7.5M
M
()
M.
M O
000
0 0 0 0 .0
M
000
()
5.0M
M
MO
1 0 0 00
M
M
00
0
2.5M
MO
M
M
M
C
O.OM
M
M
M
M
M
0
20
40
M
60
Figure 4.0 Distribution of Blue for Machine 2
12.5M
M
M
MO
0 0)
M
M
M
00 0
00
00
0
0
M
M
00
()
C
) .0
MO
000
0000
0 0 0
M
0 0 0
00 0 . 0
M
M
5.0M
00
00
0 00
C
M
M 0
()
M
M
()
()
2.5M
M
()
()
M
M
M
O.OM
M
M
20
M
M
M
0
40
60
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Figure 4.1 Distribution of Red for Machine 2
43
()
10.OM
Number
of
Red
()
00
0000
0 .0 00 0
Number
of
Blue
.
0 0)
.
7.5M

Test
Description
Approved For Release 2003/04/18 : CIA-RDP96-00787R000200150011-4
Scoring
1
S3
S4
S5
S6
Halstead Category
Test
Nonverbal test requiring abstraction of conceptual relation-
ships. Score: Total errors.
7
14
33
26
6
28
Tactual Performance
Test
Requires placement of 10 geometrically shaped blocks in
their correct locations on a formboard while blindfolded.
Separate RT, LT, and bimanual trials.
Score: Total time (min.).
16.4
11.8
7.7
7.7
11.4
6.9
Speech Perception
Test
Discrimination of non-word speech sounds.
Score: Total errors.
4
2
0
2
5
3
Seashore Rhythm Test
Discrimination of nonverbal rhythms.
Score: Number correct.
27
25
28
29
26
29
Finger Tapping Test
Measure of finger oscillation rate for 10-sec. period, both
RT and LT hand trials.
Score: No. taps/10 sec.
RT/LT
53/50
RT/LT
53/49
RT/LT
48/47
RT/LT
54/53
RT/LT
47/47
RT/LT
48/43
Trail Making Test
(Part A)
Requires connecting numbered circles in order from 1 to 25.
Paper and pencil task. Score: Total times (sec)
40
16
18
19
30
27
Trail Making Test
(Part B)
Requires connecting alphabetic and numbered circles by
alternating 1->A+2->B, etc. Score: Total time (sec)
56
50
55
50
54
53
Knox Cube Test
Measure of attention span and immediate visual memory.
Score: Number correct.
13
14
13
16
17
17
Raven Progressive
Matrices
Nonverbal intelligence test involving spatial matrices.
Score: Number correct.
39
53
49
55
60
54
Verbal Concept
Attainment Test
Requires abstraction of verbal conceptual relationships.
Score: Number correct.
22
24
27
23
21
24
Buschke Memory Test
Requires learning a 20-word list in a maximum of 12 trials with
repetition of words omitted after each trial. Score: Max. no.
words correctly remembered; List: no. words consistently remembered
Total:
14/20
17/20
18/20
19/20
20/20
20/20
List:
8/20
14/20
11/20
16/20
15/20
(8 trials) (7 trials)
16/20
Grooved Pegboard Test
Requires insertion of 25 pegs in their holes in a pegboard. Both
RT and LT hand trials. Score: Total time (sec).
RT/LT
76/74
RT/LT
69/70
RT/LT
58/67
RT/LT
59/67
RT/LT
RT/LT
48/50
72/70
Spatial Relations
Subtest of the PMA
Requires mental rotation and identification of figures
rotated in 2 dimensions. Score: no. correct - no. errors.
-
60
52
Gottschaldt Hidden
Figures Test
Requires tracing outline of simple figure hidden within
lines of more complexApproved For Release 2003/04/18. : CJA-BDP.96-00787R0002001500Av4-
-
v. good
outst. outst.