MEDICAL DEPARTMENT 
U.S. NAVAL AIR STATION 
A • * 
QUONSET POINT, R. I. 
16 April 1945 
THE EFFECT OF NIGHT VISION TRAINING (EVELYN TRAINER) 
ON U.-.S. NAVY RADIUS PLAQUE ADAPTOMETER SCORES. 
Research Project: BuMiS No. X-558 (iiK-289-p; 
Report No. One 
X 
Principal Investigator; J. Urlansky, Lt.(jg) H(S), USNR 
Collaborator; w* Dana, Captain, (MC), USN 
APPROVED; 
Ini* DANA 
Captain (ivIC) TJSIJ U.S. NAVAL, AIR STATION 
QUONSET POINT, R.I. 
OBJECT;, To investigate tne effect of the Evelyn two-dimensional night 
vision trainer on scores on the Navy Radium Plaque Adaptometer 
(RPA) Test. 
fiETHOD; 'To compare with a control grour tne percent failing the RPA 
Test after training. 
SILiiiARY AND 
0 ON CL»U S1 ON S : 
Following one session on the Evelyn two-diaensional night 
vision trainer, a group exhibited 16,6/6 failures compared 
to 18.4/o failures in a control group on the RPA test. 
nhqn the test immediately follows training, there are less 
failures (14,9/oj than when tnei»e is a delay of one week 
(19,6/oj, 
men who have previously passed the RPA, are then Evelyn trained 
and subsequently RPA retested still show 8,3/b failures. 
In a small group, 13.8/o of previous passers fail the RPA 
upon retest. None of these men were Evelyn trained. 
nhile few of these findings are statistically reliable, tne 
observed tendencies are consistent and in the same direction 
when measured by several indicators. The same tendencies are 
observed even after those individuals getting a score of 0/20 
are arbitrarily excluded. BACKGROUND: This study was designed to discover whether the number of men 
ordinarily failing the Navy Radium Plaque Adaptometer Test (Device 9-B-4) 
might be reduced if this test were preceded by training on the Evelyn 
Kighu Vision Trainer (Device If training improves group performance 
on the night vision test, one might then prescribe training to limit the 
number of failures, nhile tne incidence of failures could be limited 
arbitrarily by brightening tne Plaque or by liberalizing the performance 
requirements on the present test, training might insure that personel 
approacn tne test with a more equal and relevant background of experience. 
Should this be confirmed by evidence, the test would then intrinsically 
be more uniform for all and performance tnereon a more significant indi- 
cation of night vision ability, 
LAPuKLn..NTnL PROCEDURE; During tne course, of this experiment, men who appeared 
for night vision training were required also to be tested for tneir night 
visual ability. In some cases tne test immediatly followed the training, 
while in others tne test cane after a week’s delay, Tnis permitted an 
estimate not only of the effect of training but of the duration of such 
an effect as well, A11 subjects were questioned concerning previous night 
vision training and testing they might nave encountered. As a result of 
tnis information, some observations are possible concerning the effect of 
a previous RPA test on subsequent RPA scores. None of the subjects had en- 
countered any form of night vision training prior to this experiment. 
«hile it mignt be of great interest to measure the effectiveness of the 
Evelyn trainer on nignt visual performance—and this is partly done in 
tne present study--tne principal aim was to ascertain whether the RPA 
test results migive be influenced by some training procedure. Accordingly, 
tne percent normally failing ins RPn may be taken as the standard against 
which comparisons may be evaluated. Information based on the normal test- 
ing of 4477 men at this Station was available and used for this purpose. 
Both the Evelyn training and the RPA test were performed routinely on the 
men involved in this study. Adequate dark adaptation was insured by naving 
the subjects wear red goggles for twenty minutes and then spend ton more 
minutes in complete darkness before being tested. The Evelyn training ses- 
sion, wnich lasted about an hour, demonstrated tne need for dark adaptation, 
peripheral vision and scanning procedures. 
In all instances, tne Evelyn training proceeded the night vision test and, 
indeed, some men were led directly from one to the other. These constitute 
our immediately tested” group. Some other men were tested a week after 
the training session. 
Among our subjects were some men for whom tnis experiment meant that they 
were taking the RPA test for the second time and these have been treated 
separately in the data. 
♦ ♦♦Thanks are due Lt.Comdr. N. B. Combs, (mC), USNR, for his cooperation 
on this study. RESULTS; The principal findings are summarized in Table 1, A total of 
648 men were tested. 
The performance of tnese subjects may be assessed by comparing them 
with the records of groups that have been routinely run through the 
RPA at this Station. During the period of June 1944 to February 1945, 
4477 men were tested and, of these, 18,4/o failed to reach the standard 
Navy performance. (Fig l) The median score for the whole group is 
10.28 correct out of 20. This figure approximates that reported by 
the U.S, Submarine Base at New London, Connecticut, where 17.6/o men 
fail in a sample of 5750 (l). The present authors have collected 
evidence which suggests that the number of men failing the RPA test 
varies with the seasons of the year, the significant factor probably 
being exposure to sunlight; this will be tne subject of a separate 
report. The matter is pertinent to this report in setting up a 
proper control against whicn tne experimental effects may be estimated. 
A sample of 1297 men were tested here during January and February 1945, 
while out data were being collected, shows that 18,7# failed. As one 
would expect, this figure is not significantly different from the 
average for the whole group; (the chi-square test yields p = .77). 
One experimental group of 434 men showed 16.6/q failures on the RPA 
after having been trained, instead of the normal 18,4#. Its median 
score was 18.61 and this is somewhat higher, i.e., better, than 18.28 
in the control group, (Fig 2} Tnis information becomes more meaning- 
ful if analyzed to reveal the influence of a time delay between train- 
ing and tne test. v<hen the test immediately follows the training 
session, as it did for 276 men, there are 14.9# failures and a median 
of 13.69. After a delay of one week for 158 men, the failures rise 
to 19,6/0 and the median drops to 18,44. None of the percent deviations 
from tne control measures are statistically significant by the chi- 
square test, tnough tne difference achieved by the immediately tested 
group approaches this -with a ,fp** equal to .15. Such an indication may 
be taken to mean that if there are any effects of training as can be 
measured by tne RPA, these are short-lived indeed. 
By far, the best performance was turned in by a group of 134 men, who, 
prior to this experiment, had once passed the RPA test. When these men 
were tested again, in this case immediately after the Evelyn training, 
only 8,3# failed and the median score is 19,03; chi-square shows the 
difference to be reliable with a np** equal to ,01. This drop in the 
number of failures may be due either to the RPA retest as such, to the 
selection of previous passers, and to the Evelyn training. A suggestion 
as to the relative involvement of these factors may be gained by inspect- 
ing our final group of 80 men, all previous RPA test "passes", who were 
retested though not Evelyn trained. Here there are 13,8# failures, a 
drop from the 18.4# normal group as substantial in magnitude as that due 
to training (14.9# for one sub-group, 16.6$ for the entire group), with 
both groups identical in all respects except for the matter of Evelyn 
training, the difference between 13.8# and 8,3# failures may well be an 
indication of the influence of Evelyn training; while the difference 
between 18,4# for the normal group and 13,8# for tie retest group may be partly due to the retest. By far the more disturbing fact, however, 
is that 13.8/a of a group that once passed the RPA may fail on retest. 
As reported above, the variations in both the percent failing and the 
median score may be generally interpreted to point to the sane tendency. 
To a lesser extent, this is true as well for other indications of per- 
formance that may be found in tne percent achieving a perfect score (10/10) 
and in the percent failing to see anything at all (o/20), (Fig. 3) Cumu- 
tive percent curves in Fig. 4 also show the separation between the several 
groups, 
DISCUSSION: 
Since few of tne observed differences in percent failing are reliably dif- 
ferentiated from the control measures, some caution must be exercised in 
interpreting these results. Not only are the sanples relatively small, but 
the RPA, as an experimental indicator, does not mirror all the functions 
one is likely to train in night vision. One night also prefer the RPA to 
have a higher test-retest reliability. However, there are certain clear 
tendencies in the data, and another investigation-with larger groups 
might yield not only direct verification but even statistical reliability 
for differences of the same magnitude. 
with these cautions in mind, one may assert that Evelyn training only 
slightly (l,8;o) diminishes the incidence of failing the RPA, though the 
effect is larger (3.5/o) when the test immediately follows training. This 
finding is a disappointment should one propose use of the Evelyn to reduce 
sharply the number of failures. A broader view may be that the substance 
of Evelyn training can improve tne lower threshold of night vision, one 
may assume that the effective carry-over to the test is some knowledge 
about peripheral fixation and possibly a greater motivation to succeed. A 
single session in the two-dimensional trainer scarcely affords much prac- 
tice in scanning or peripheral fixation, and additional training may be 
necessary to instill these as working habits. As measured on the RPA, the 
effect of training wears off rapidly, which fits in with these observations 
Training more likely to decrease RPA failures would have to provide ex- 
perience relevant to the test, and this information already exists. Thus, 
the New London Sub Base reports that when failures are retested, only 5% 
of the ppulation still fail after three tests, (l) 
It may well be that the performance required on the RPA is so simple and 
basic that, except for the matter of fixation, it is only relatively sus- 
ceptible to improvement by training. For this purpose, direct practice in 
fixation, repeating the test and greater emphasis on fixation by the test 
operators would result in a more uniform performance. By the same fact, 
however, one should not expect it to give the best all-around prediction 
of night look out ability. 
The question as to just what the two and thr.e dimensional trainers achieve 
except for demonstration, remains unanswered and is not properly the concern 
4 of this study. However, such an investigation is imperative for testing 
the present effectiveness of the devices, for developing an inclusive 
curriculum, for establishing practice procedures, and for instituting a 
level of competence whicn personnel must reach before being qualified for 
night lookout duties. 
Indeed, the least failures (8.3/o) are among those who, having previously 
passed the RPA, have now been trained and retested. Selection, training 
and retest all contribute to the reduction, the intermingling of which may 
be partly judged by the percent of passers who fail on retest alone (l3.8/o) 
without benefit of training. One possible flaw in these data is that it was 
impossible to check in all cases the subjects’ claim of a previous passing 
performance on the RPA* Let us- note at least, that the group one would ex- 
pect to be most favored, with selection, training and retest, still exhibits 
8.3/0 failures. Evelyn training an unselected group could hardly reduce 
beyond that figure the percentage of failures on the first RPA* 
On a purely statistical basis the evaluation of these results must be 
tempered with caution. Five out of twenty-one different possible scores 
constitute a passing performance, while the remaining sixteen are called 
failures. The resulting frequency distribution is hardly normal and this 
is a function, mainly, of the failure to discriminate effectively among those 
who pass (approximately 82/o of the population/. To a lesser extent, another 
contribution to distortion of the curve is made by a segment of the population 
(approximately 8/0) who probably feign night blindness and get scores of 0/20, 
This casts doubt over any statistical procedure based on the expectation of 
a normal distribution. 
Since none of the observed effects in this study may be said to be unduly 
large, the existence of a large group of individuals who fail in the 0/20 
category, i.e., by failing to guess right on the test not once in 20 trials, 
constitutes a hazard for any interpretation. On a chance basis, there are 
three ways out of four than an individual may be wrong on any single trial 
and 3 chances in 1000 that this could occur on 20 straight trials . Some- 
thing pther than chance must be operating when 8/b rather than an expected 
.3/b of the group get a score of O/20, men examined on the RPA may feel that 
a failure on the test will save them from going to sea and that there is an 
advantage in not being able to see anything at all. The method for achieving 
this is naive when the individual for 20 trials calls the letter UT" in a 
position directly opposite from that in whicn it is placed. Charting the 
dark adaptation curve or scotopic perimetry might detect the fraudulent. The 
presence, however, of such a group in an experiment might easily mean that 
the observed results are due less to the influences being studied than to the 
capricious nature of these examinees. In the 9 months during which the RPA 
has been employed at this station, the number of men getting this score has 
varied from 0 to 12.0/o, with a mean of 8,5/o for the total group, 
we assumed, therefore, that most of these men were probably feigning night 
blindness and desired to evaluate the. data in a way to be free of any in- 
fluence these men might contribute. A limit of 2/0 failing completely (o/20) 
was established, and those making up more than that percent of the population in each group arbitrarily excluded from the data, which was then reworked 
and presented in Table 2. The main finding is that the tendencies originally 
observed still remain and we may, with some confidence, assert that the data 
is not an artifact due entirely to the chance composition of the 0/20 
groups. (Fig. 2, 3). 
REFERENCES: 
(1; A report on the night vision testing of 5750 men. Medical Research 
Dept., U.S. Submarine Base, New London, Connecticut, 12 Aug 1944. 
6 TiiBLxii I 
Su.ri:aary of Data 
N 
/ 
/O 
Fail 
Median 
Score 
/ 
10/10 
scores 
/° 
0/20 
scores 
P value 
test for 
against control 
Control - Quonset Ft. 
4477 
18.4 
ia.28 
41.4 
8.5 
Evelyn trained, RPA 
434 
16.6 
18.61 
45.2 
7.6 
.35 
tes’ced .mole group 
Evelyn trained. 
tested immediately 
27S - 
14.9 
18.69 
46,0 
8.0 
.15 
Evelyn trained tes- 
1*58 
19.6 
18.44 
43.6 
7.0 
.65 
bed one week later 
Previous RPA pass. 
Evelyn trained. 
134 
8.3 
19.03 
51.5 
5.2 
.01 
RPA retest 
Ho Evelyn training. 
80 
13.8 
19.11 
56.3 
7.5 
.25 
:A a test and retest 
* -men this value reaches ,0b 
or 
less. 
then differences 
as large as 
this due to chance factors coul 
.d have 
arisen 
only 5 times in 100 
and the difference may 
then 
be 
considered reliable. TABLE II 
Summary of Data 
Recalculated after limiting failures 
in the 0/20 category to 2/0 of the population 
N 
% 
Fail 
Median 
Score 
i—• 
o 
1—1 
o 
o/zo 
P9value 
test 
against 
control 
Control - Quonset Pt. 
4188 
11.9 
18.55 
44.2 
2.0 
Evelyn trained, RPA 
Tested, whole group 
410 
11.7 
18.83 
47.8 
2.0 
.88 
Evelyn trained, tes- 
ted immediately 
260 
9.6 
18.92 
48.9 
2.0 
.25 
Evelyn trained, tes- 
ted one week later 
150 
15.3 
18.67 
46.0 
2.0 
.20 
Previous RPA pass, 
Evelyn trained, 
RPA retest 
130 
5,4 
19.06 
53.1 
2.0 
.03 
No Evelyn trailing, 
RPA test and retest 
78 
9.2 
19.16 
57.7- 
2.0 
1.00 
8  
£?/ST~fZ/Bur/O AJ O*:' P.P.A. 7~ L/OA/S:t^O/A/T 
k/’oa/s /94+ — 3 794 S' ( A/ * 4+77) 
/P/*4 ~3cokb: 
// // 
P+7L. 
(/s.+ jo) 
/=m/<SrV*.B: / 
" jt A CAlCaiATBO PBBCBA/T 
ABTBR 
APB/TBAB/l r I /r?/r/M<r 
TB05B J5COK/AJ(j °/zo 
TO Z</o OB -r~//£ 
B>OP(JLA T/OA/ 
k to 
«s 
k 
x £ 
£ 
<* k 
\ 
5 I 
* § 
i 
k <s 
k « 
*J K 
k 
/P 7~/r*57~ - w 
croA;/e£rc.T&£> 
/A/£*/<Z£rS 
2. 
/*&kc£ast 
&&Tr/*s<f 
J&JF3T 
'°//o 
Wo/?sr 
THE 
MfD/AU 
j5COK£. ct/s>7 u tsg- /e 
CL//iULA T/VE PERCE OURI/B6 
ppfspopppy/ot/s 
£y£-jLp*y ~m*/A/t=rp ? 
ppp 
A/=t /34 
4/0 pygi r*/ 
JPPP 7~/FsT 4 
O/vAy 
a/* so 
P'VP’l YA/ r~P>/?/V/Srp 
PP/J res p 
/P 434 
Z/oP/rf/9/. PP/f OASLy 
P*r. 
M ??77