ENGINEERING RESEARCH INSTITUTE THE UNIVERSITY OF MICHIGAN ANN ARBOR Final Report MODEL SIMULATOR STUDIES OF THE VISIBILITY OF MILITARY TARGETS AT NIGHT Charles E. Hamilton Vision Research Laboratories ERI Project 2699 THE HUMAN RESOURCES RESEARCH OFFICE GEORGE WASHINGTON UNIVERSITY SUBCONTRACT NO. HumRRO-1-003 CONTRACT NO. DA-49-106-qm-1 WASHINGTON, D.C. August 1958

This report presents the results of work done under the administration of the Engineering Research Institute, whose name was changed to The University of Michigan Research Institute on July 1, 1958.

The University of Michigan * Engineering Research Institute 2699-1-F TABLE OF CONTENTS Page List of Figures iii List of Tables vi Summary vii I. Introduction 1 II. General Experimental Conditions 5 III. Experimental Results 15 IV. Discussion of Results 25 References 33 Appendix A 49 Appendix B 53 ii __________________________

The University of Michigan ~ Engineering Research Institute 2699-1F LIST OF FIGURES Number Tit le Page 1 The model simulator. 56 2 Experimental room and equipment 597 3 M-48 tank scale mode1. 58 4 M-59.A scale model. 59 5 ~ Anti-tank 8guIn ad crew scale model, 60 6 Map of original target positions during exploratory experiments. 61 7 Map of final unaiform'target field and target positions. 62 38 F~Map nf nonuniform target field and target positions. 63 9 Results for MH48 tank. observed on nniform t errain under starl(ght illumination, sIhowing relation between freqguaency of response data and information measure M o 64 10 Results for, M-48 taLnk and M 59 APC. obsetrved Hon unlf tom te' rin under L stalght i, lu.uinationo 65 11 Results gfor M-48 t*ank observed on uniform terrain under moonl!ight il 11u mnation o 66 12 Re$ulSg for M-59 AiK observed on uniform terrain t nde moonlight ila mina ion, 67 13 Resullt for antitta nk grn and crew' observed on. uniforn ierrain unde r moonlight illulnat'iono 68 14 Results for KM48 tank9 obgerve. on non-unifor,- terraiV"'n in midfiteld position unader moonlight illuminatiomn 69 15 Results for M0 48 tanky &bs~en.ed on non=uniform terrain in tree position under 4-,sec searchlight illulmination along obserer's line of sight. 70 16 Resnalts for MA,8~ t amk9 obsertved on non-,:niform te'r ain in tree position ilusminated by searchlghit at differing ditspac*e-, nts from ebserver. 71. 17 Results T for M48 tank, obc served n diffe en t ntonunTfomin f ijght i rLumi alto on 72 iii iT'

The University of Michigan ~ Engineering Research Institute 2699-1-F LIST. OF FIGURES (Cont'd) Number Titl e Page 18 Results for M-59 APC, observed on non-uniform terrain in tree position under 4-sec searchlight illumination at differing displacements from observer. 73 19 Results for M-59 APC, observed on non-uniform terrain in upper field position under differing durations of searchlight illumination. 74 20 Results for M-59 APC9 observed on non-uniform terrain in tree position illuminated by searchlight at differing displacements from observero 75 21 M-48 tank on uniform terrain (mid-field position) under moonlight illumination. 76 22 M-59 APC on uniform terro rT (mid-field position) under moonlight illumination 77 23 M-59 APC on uniform tsefrrain (mid-field position) under illumination by searchlight 75 yards right of observer. 78 24 Anti-tank gun and crew on uniform terrain (mid-field position) under illumination by searchlight 75 yards right of observer. 79 25 M-48 tank on non-uniform terrain (mid-field position) under moonlight illumination. 80 26 M-48 tank on nonwuniform terrain (upper field position) illuminated by searchlight 75 yards right of observer. 81 27 M-48 tank on non-uniform terrain (upper field position) illuminated by searchlight 150 yards right of observer. 82 28 M-59 APC on non-uniform terrain tupper field position) illuminated by searchlight 75 yards right of observer. 83 29 M-59 APC on non-uniform terrain (upper field position) illuminated by searchlight 150 yards right of observer. 84 30 Target and terrain luminances for the three scale model targets when illuminated by searchlight along observer's line of sight. 85 iv

The University of Michigan * Engineering Research Institute 2699-1-F LIST OF FIG.URES (Cont'd) Number. Title Page 31 Target and terrain luminances for the scale model vehicle targets when illuminated by a searchlight 75 yards right of observer. 86 32 Candlepower distribution for simulated searchlight beam using cotrector slide. 87 ______________________ V....._______________

The University of Michigan * Engineering Research Institute 2699-1IF LIST OF TABLESS Number Title Page I Results for M48 tank on uniform terrain, observed under starlight illumination 34 II Results for M'H59 APC on uniform terrain, observed under starlight illumination 35 III Results for M-48 tank on uniform terrain, observed under moonlight illumination 36 IV Results for M-59 APC on uniform terrain, observed under starlight illumination 37 V Results for anti-tank gun and crew, observed on uniform terrain under moonlight illumination 38 VI Frequency of response results for anti-tank gun and crew observed on uniform terrain under searchlight illuuination (observer C H,) 39 VII Results for M-48 tank on non-uniform terrain observed under moonlight illumination (mid-field position) 40 VIII Frequency of response results for M-48 tank on nonuniform terrain under searchlight illumination (upper field position) 41 IX Frequency of response results for M-48 tank, observed on non-uniform terrain under -searchlight illumination (tree position) Observer C.H. 42 X Frequency of response results for M-59 APC, observed on non-uniform terrain under searchlight illumination ( upper field position) 43 XI Frequency of response zesponse results for M-59 APC, observed on non-uniform terrain under searchlight illumination (tree position) 44 XII Frequency of response results for anti-tank gun and crew, observed on non-uniform terrain under searchlight illumination (upper field position) Observer C oHo 45 * vl.....~~~~~~s

The University of Michigan ~ Engineering Research Institute 2699-1-F A progra of experiments iS eported in which visibility distances for military taSrgets have been assesseed uslng a Blle-model simulator., Targets were observed along ground path under simulated natural and Artificial conditions of nightsti me illuminationo The experiments were concerned with both detection and identification of the targets~ The tsrgets were selected to represent different classes of military targets and included a tank, armored personnel carrier, and anti-tank gun and crew. These targets were viewed under starlight, moonlight, and se'chlight conditions o They were always located'in unconcealed positions and, in d7iffeTent expe'iments, on relatively uniform and noan-aniform terrain. Their visiLility und "e ear chlight lIumination was studied ander a wide range of special condition relating to searchlight 4duration displacement ftom observer, and flicker. When the tarxgets were on uniform terrain' detection distances obtained under $tarlight illumination were about 190 yards for the tank and APCO The anti-tank gun and crew Could not be seen at the miniram distance it was possible to uses about 100 yardso Under moonlight illumination, the tank was reasonably detectable at between 900 and 1000 yards; the AFC was about as vliable or a little more so; 500 yards represents the detection distance for the anti-tank gun a nd crew. Identification distance under these conditions is estiCmalted at about 600 yards for the tank and APCo When the tank was located on nonn-uniform terrainP its detection distance was reduced to about 060 yard0s which is about the maximum identification range already notedo With searchlight illumination~, the vehicle targets, when on uniform terrain, could be detected at the maximu range of 1500 yards0 The anti-tank gun and crew were visible to about 1000 yards under the same conditions o Targets located on non-uniform terrai and viewed under searchlight illumination were detectable in a complex way as a function of inmmediate background, duration of searchlight illumination~, and searchlight displacement from the obserVer.o The poorest visibility ocCurred when the targets were against a tree bac kgrournd for short durations of illumination, and with the searchlight not displaced from the observer. Under these conditions, again, the detection.rage was about the same as the identification range noted earlier Attention w~a given to detet ining when pos$sible, the stimulus factors underlying the vi$ibilty of the t rgets o I this Tegard, photometric data allowed scme detemination of correspondence of results with prediction from mEore basic vis al dte-sion dta~. _ hotometric data were alo used to relate t aemulator co n.4ixtons$ to a tua( field 4ondJ tiLonss | __ __ __ _ __ __ __ _ __ __ __ _ __ __ _ i _ __ __ _ __ __ __ _ __ __ __ _ __ __ _

The University of Michigan ~ Engineering Research Institute 2699-1-F in order to evaluate the degree of correspondence. Also, certain of the experiments. employed conditions similar to those used in a field study conducted elsewhere, and provide a basis for comparison of performance data. viii _ _

The University of Michigan ~ Engineering Research Institute 2699-1-F I, INTRODUCTION The present report suri-zes- the experimental studies conducted under Engineering Research Institute Project 2699, established in terms of Subcontract 1003 with the Human Resoutces Research Office, George Washington University, WashingtonB D. C. These studies were part of a program intended to supplement the continuing program of the HumRRO staff of the UoS. Army Armor Human Research Unitt at Fort Knox, Kentuckyo In general terms,, the model simulator studies have been concerned with the determination of visibility distances for military targets under certain conditions of night time illumination. The studies are intended to provide, also, a basis for better specifying and understanding l the stimulus factors influencing the visibility of targets under such conditions. Since the specifications of target, terrain, and illumination conditions were intended to relate closely to problems of special interest to the Human Research Unit at Fort Knox, the establishment of these conditions was carried out in close coordination with personnel of that Unit'. To achieve this coordination, the writer visited the Human Research Unit at Fort Knox during November 6 and 7, 19579 for the purpose of conferring with Dr. Howard McFann and members of his staff. Extensive discussion was held with Dr. McFann Dr. Norman Willard Dr. Nicholas Lewipa Dr. Ed Stark, Dr. Fogel Clark, and Dr. Al Kraemero The discussion had as its objective to make the stimulus conditions of most direct interest in relation to the studies being conducted at Fort Knox. Agreements (to be listed in detail later) were reached concerning the target, terrain, and illumination conditions for the model simulator studies, as well as the general psychophysical procedures which would be followed. Subsequently to the conference described above, correspondence between the writer and personnel of the Human Research Unit served to clarify problems and maintain a close relation between the model simulator studies and their counterpart stVdies at Fort Knox. Visits by Dr. Willard in March and May, 1958,'and a visit to Fort Knox by the writer and Mr. Carl Sermmelroth to observe the field study, Armornite V, during June, 1958, further aided our objectives o The target, terrain, and illumination con oditions, and psychophysical procedures agreed upon at Fort Knox in November, 1957, will now be outlined, With one exceptiong noted in the context of the outline, we were able to establish all the required conditions and obtain observations over the wide range shown. General interest was confi:ned to night time illumination conditions and the observer's task was to be restricted by the following features: (1) The observer has (some) knowledge of the type of target to be detected and its lateral displacement (2) the target is considered only an "enemy" and determining time for detection is of importance. In addition to item (2) above, recognition responses were to be included in the context of the staudies.....________________________________ 1 ___________1

The University of Michigan * Engineering Research Institute 2699 ~ — F The simalate, field iat.i2ti'on was conceived as that peculiar to the tank platoon This nsiderat ion ditted that the maxirnm displacement of targets should be r ttrite d to th qu ival ent of O 200 yiards, andSp as well, that searchlight -mggles o(f ill m1 nation an d~isplacement from the observer should be conosstent whth tihos t.ypi l of the tank platoon -nder operational condition o Cornsideration was given to the factor of atmosphere in the model simulator situation. Since relatively great interest was expressed in connection with art ificia l {earchlight. ) ilyruination effects, it was evident that a technical proble aexist6d In $imtlating the optical effects of the atmospherea in the abs orpio and a catter of light two pos$ible techniques msy be *tuo The tfirst involves generating artificial atmosphere by means of as8ing water droplets a pd opaciti es to the actual atmosphere srrTounding the mod0el. The second uti lizesI instead, a "veiling luminance". This'is accomplished by placing a parti'aly-silvered mirror before the obsr2vers8 ~B esvnd t-eSf l&nc t lKg wltth Lt n e of uni't4for lTninanxe superi Lmposed over the obserer's view of the mode Lo The late procedxmeE is completely adequate to simul1 ate maspher'ic. effects occturri.g between the military tzget and tbhe obse vero Hiwever Pit does not simulate other effects such as the.pperance of th te searchlight beam in passing through the atmosphereo Consequently, the f irst pr(cedaue outlined appeared to be the only acceptable procedae t. hough its feasibility could not be attested. The use of artif icial fog,, wh1ch would depo$it water droplets on the model al$so, appeared cosmplicated d'6e to the physical dimensions of the experti menttal room. PreviouS work had been concerned with simulating dense fog, and the production of relatively light fogs ln this manner still remained to be investigated0 Si.-nesNtu, it appe red achieving such conditions posed grave technical problems, it 7as agee to centgct the sttuies initially ~ i th no ~elay for et the stud i i ial ly th no de fo or th ptoe s of introucing. scale'd atmosphere o If later such conditions could be obtainedT they should be added, but, in any event$ the results olf" studlies wee to be rlated to atmosphere effects in final aThe s~pao,. LJ.. condtitssns to be simulated ere agreed upon andI are shoa, below:o A.o Tag. The folo I (9ing a rgets were to be $d singly in a given eperi mental st'atiuono 10 ed:.um U.oSO Tank (M48), 2o Armored personnel eathier (M-59) 3 k nti-tank gun w th ctrew These t*gats we ao be pos i ttior( hptAon d to the obster vetr, not d-Eg iu or c:asoflaged.. The ~ti-tak g8n (ad ctoer were t.o be aranged in the open as, carew wor~king ar d thr gun,'agettitng into po$itionvwo

The University of Michigan * Engineering Research Institute 2699-a1F B. Terrain Backgroand, 1o Homogeneoue ( suh as wheat field) 2. Heterogeneous (bra'sh tres etc o) C. Illumination 1 o Natural ao Dark nightg with stIlight only bo Brig tht night, with fl moonlight 2o At ificia l o One or more 18i nch tank mounted searchlights b0 Flare It should be noted at this pot tt hadue to limitation$ of times and the relative mpota c assigned b ths Human Rese arch U nit to item 2 a, searchlight illtumnation, the studies to be reported did not encompass item 2 b, flare illurinatlO'no The cond0it.tons for searchlight illuminati.on were listed in more detail than show above. These special conditions included not only a single s erchlig.t at dfifering displacements from the observer but alsco two searchl igh1salterneately lluminating the target (the latter being under contin s oc illuminationl) In addition, intermittent single searchlight ill' ~n nation ws, to be included, and the illumination'schedules for this anC the foregoing coondi on were indicatedo The general. psychophygica1. procedure agreed upon was to have the observer make ob ev8rvion at a fxedd di tance of each target poSitioned on each terrain background under each of the illumination conditions listed. The procedure was ntended to lea to freqluerny-ofseeing data for each experimental condition, The ob se-rvers response was to be timed with respect to detection, and, in ad(ditionr, he was to attempt to make a "l.lass'i identification follow ing the detection,of I t rgeto In the studies rpo te d in the follodng sections we Ettnigpted to adhere as c los as closely pssible to the agreed upon conditions. In addition, in onnection with other phases of tkh subcontract, we attempted to relate our conditions as closely as possible to actual field conditions. Information concerEning such interrel&attionship with field conditions will be given later in this reporto And, finallay,, within the framwork of the foregoing specific conditions, we atterlmpted to utilize the advantages of the model simulator situation (repeatabili+y of particular conditions, etc.) to gain as much insight as possible into the observer$s task, effects of practice, and stimulaus factors of importance in a general sense as well as determtining the particularized target visibility distances for the specific clonditions studied To pres$ent ors fiLd:iS84 g simpl y as possi$ble the material has been organized into three sectiona which follow Section II cont/rns a

The University of Michigan * Engineering Research Institute 2699-1-F description of the special equipment and general experimental conditions used. Section III contains the results for the series of experiments conducted within the framework outlined in the present section. In section III attention will be given primarily to presenting the particular findings for each condition of observation described. In section IV the results of the different experiments will be interrelated in a more general manner and attention will be given to exploring more fully factors underlying the form of the results. 4I

The University of Michigan * Engineering Research Institute 2699-1-F II, GENERAL EXPERIMENTAL CONDITIONS The model studies were conducted using a three-dimensional terrain model, at a scale of 108:1, measuring approximately 20 feet by 20 feet. This model represents a sample of actual terrain with hills, meadow land, highways, a river, and a town. The model is complete with surface detail. Figure 1 shows an over-all view of the terrain model, The terrain model was developed initially under sponsorship of a tri-service contract administered by The Signal Corps (Project MICHIGAN), for use in earlier studies relating to the visibility of military targets along ground paths. The model, as well as certain other equipment, was used by permission of Project MICHIGAN for the present studies. The terrain model was located at one end of a room approximately 60 feet long by 30 feet wide, and 20 feet high. The walls and ceiling of the room were painted black to reduce stray interreflection of light. The room itself was made highly light-tight to afford good control over the low levels of illumination intended. All personnel taking part in the experiments and all equipment used were in this single room, Blackout conditions were maintained during sessions except for the special illumi nation introduced. To isolate the observer from the activities of the experimenter, he was seated on a theatre chair in an enclosed observing booth mounted on castors. When the booth door was closed, the observer's vision was restricted to viewing the terrain model through a cutout window 30 inches wide by 20 inches high centered 20 inches in front of him. A shelf imediately in front of the observer supported a chin cup and forehead rest assembly which was intended to make for consistent head positioning. The observer's eyes were 46 inches above the floor. The shelf also held an intercom unit and a pushbutton box which the observer used to qconal his detection responses. The use of this equipment will be'described later. Because the observer's booth was positioned at varied distances from the terrain model, several different black cardboard cutouts were available to fasten over the window of the booth so that in each case the observer saw only the terrain model. The model was seen approximately from its left to right extremities, and seen in the vertical dimention only from the lower edge of black cuLtain below the model to a point in the!"sky" just below the ceiling. With these restrictions, the observer was prevented from seeing any of the sources of illumination or the activities at the experimenter's desk on the left side of the room near the modelo Figure 2 shows the experimental room in most of its details. At the right is shown the side of the observeA booth facing toward the terrain model in the background. At the left, against the wall, is the experimenter's desk containing the master intercom unit and a one-hundredth second electric clock timer associated by a special circuit with the observer's response box, The experimenter's desk was illuminated by a shielded red lamp, which under dark-adapted conditions was adequate for the activities required. (It should be added that during sessions, the...~~~~~~~

The University of Michigan * Engineering Research Institute 2699 -'F experimenter used only a red filtered flashlight to guide his way to and from the model itself to reduce the possibility of accidentally affecting the observergs state of dark adaptation.) Certain other features seen in Figure 2 are of interest, First, the long white line seen on the floor parallel with the edge of the booth is a guide line for positioning the observer, An indicator, seen projecting from the lower left edge of the booth, was at the same distance from the target as the observer's eyes, and by appropriate pos'ti.on.ing the booth indicator at distances marked on the guide line, any desired distance from a given target position could be obtained. It may be seen that the. booth faced the terrain model from an angle of about thirty degrees. This was necessary because the model itself was lowest in ts center front and sloped upwards in all directions from this point. By positioning the booth as was done, the observer viewed the expanse of the left side without obstructions in the foreground. This portion of the terrain sloped gently upward going away from him, and termin ated in trees at the skyline on the left leading to the hills on the right~ For part of the sessions., the main part of this area was uniform (as may be seen in Figure 1) and for the remaining Sessions it was made non-uniform with bushes and trees (as may be seen in Figure 2). The details of the terrain surrounding a target will be found in the next section, Attention should be called to the special equipment also seen in Figure 2o On the left in the foreground, on its tripod, is a special photoelectric telephotometer used in these studieso The use of this instrument, developed by Mr. Benjamin S. Pritchard, will be described later. Between the telephotometer and the observer's booth is seen one of the projectors used to similate searchlight illuminationL on the terrain, This is seen with its light shield cover off Its arrangement and use will be described later, also. In the remainder of this section details will be given concerning the target models, sources of illuminat'ion. photometry, observers, and, experimental procedures used. Target Modelss The three targets specified earlier were corn-, structed to scale (108:1) and are shown. in detal in Figures 3, 4, and 5. The M-48 tank was modified in detail from a commercially produced scale model (Authenticast) based on photographs and dimensions provided fzom the Human Research Unit. The M-59 armored personnel carrier was constructed in its entirety from this information since no cozmercially made model was available o The anti-tank gun and crew, consisting of six men, was assembled as an integral unit, The two vehicles posed an interesting problem with respect to their color, Since what is termed "OD." color ranges over about fourteen Munsell notations, it was considered imperative that our target models correspond as exactly as possible to the field tanks at Fort Knox. A swaple of paint forwarded from Fort Knox by Dr. Willard was used to secure this control, The actual "OoDo." resulting can be described in Munsell no ation as 2,5 yellow, 2 vaslue and 2 chromao (This i$s in dis$tinction to the unmodfied authenticast color which can

The University of Michigan * Engineering Research Institute 2699- 1be described as 10.0 yellow, 5 value, and 2 chronma.) The originaf l t ank paint is a semi-gloss paint but it apparently abrades on exposure to dirt and air and the vehicles in the feld e essen6 illy 889 mag tte" in appteaz ance and quite dko. Accordingly, our -odels were spray-painted to meet these color specificnationsm The anti-tank gun anA crew were p&antel'ith similar paint for the O.D. portio ns. The f aces of the crew were approx"% mately flesh co!oo although in oorabait prob,ly they would be blacke.ned. In general, it was felt that detail of the targets should be made as accurate as possible even though under the inntened viewing conditions an observers acu:ity for such detail would be poor. Sources of Illumination: First w11 be described the me-ans of simulating natural field illuminati n, i.oeo. 1tazlighl/t, and moonlight. Ideally, to sinamate starlight, a sky dome would be required. This would provide the diffuse illumination of the ground typical of the clear r4.'gh:t. sky. By reason of practical onsidera iois n sky dome was ny approxi mated by providing hWghly diffuse illu'na ation fro the ceiling Jrtea over the terrain model. This was ac omplishe by pi iallay covering the cement ceiling beams with matte white paper. The bem were about six inches wide and eleven inches deep from the celing. They were spaced about t o feet apart. Since the uncovered portions of the ceiling were painted black, 8abou*, ten percent of the ceiling area was responsbli e for reflect ing light directed upon it from two louvres, one at each side of the model. Each louvre contained a single one candle, six volt, lamp. The lamps were shielded so that they illuminated only the ce iling area. In turn, the model was illUminated by the light reflected from the whtole cei 18ng area, and to some small extent by re-.ef lected light from the walls. A 36 inch wide whte plast$c curta in was hung from the (e:lring in front of the model. This may be seen along the ceiling in Figure 2. This served to add to the reflected light at the front of the model. 9The lamps were powered from a Variac and transformer connected to an AoCo outlet. To achieve the lowest level of illMi~nati0on, for starlight, nTuStral t filt$ of nonal denns ity 1 were poto:iode n ithin the l0'uStvres When moosnlight levels were intended, the filters were removed to allow a greater contribution of scattered light to the illumination~. Althoutlgh the pr cedu re dess-ribed d id not provide quite e6nough 31ilaIU naoton froma the front and sides to an object on the terrain, it seemed quite adeqlute in that no disceriable shadows were created and objects appeatr uniformly illuminatedo Under starlight illumination the luminance of the uniform ground in the target area (as measured along the observer's line of Ls-ight, photoelectrically) wa about 105 foot ltmberts. MooGonl ight, in distinction to s$talight, is htighly directional in character To av thS atre achieve this feature a light oue s mounted on a scaffoldisng near tBe rear of the experimental room, The tl ouv;re enclosed, a 300 watt projector lampo The light from the lamp illuminated the mod.el from an elevation of sixteen feet and at a diistaaCe of approximate y 35 feet, o o filters were uTed to aesh'es v he deAlmed level of illmi"tton io A Wratten 78 filter altered the spectralt mpos? ion of the light o achieve the visual Iequivalent of approxma t stcl y 5C00? degrees (c~o or teapera ature, and 1an aBppropta:3Po;$e WeuittLr denst8y fil a Ufr~t e t us the Ioves _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _7_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

The University of Michigan * Engineering Research Institute 2699-1-F all level of illumination-on the model. A mask in front of the filters adjusted the light pattern so that it illuminated only the terrain model directly. In use, some light was reflected from parts of the model to the walls and re-reflected to the model. In addition, as noted above, the "starlight" louvres were operated without additional filters so as to add to the diffuse light when "moonlight" simulation was intended. The direct light from the moonlight louvre illuminated the terrain in the target area at a vertical angle of about 20 degrees, and at an angle separated from the observer (to his left) of about 45 degrees. Two features should be noted regarding the moonlight illumination. First, the over-all level yielded a luminance for the ground, as seen from the observer s position, of approximately 10-3 foot lamberts. This level is quite satisfactory for the intended simulation (Ref. 1). Secondly, however, the ratio of direct to scattered light was too great in the direction of the direct source. The desired ratio would be about 2:1. Our ratio was, in fact, 840:1. This made for more sharply delimited shadows and possibly greater contrasts under moonlight conditions than would be the case ordinarily. It was not possible, however, to achieve a better balance between the direct and scattered light for this case within the time available. For searchlight illumination we simulated the 18-inch tankmounted searchlight which uses a 2500 watt lamp of new design. This lamp has a four million peak candlepower and an 8-degree horizontal beam spread when used in the tank mounted reflector housing. Working from candlepower distribution data provided by the Human Research Unit, corrector slides were made on thirty five millimeter film slides which provided the desired characteristics of beam spread from Argus slide projectors. The candlepower distribution attained is shown in Figure 32. The illumination was scaled according to principles stated in Reference 2. To determine requirements concerning peak candlepower of the simulated source, the peak value for the 2500 watt lamp (4,000,000 candles) was divided by the square of the scale factor (108 ), yielding approximately 400 candles for the simulation source peak. The objective lens of the slide projector used as a source was stopped down until the peak reading of light projected through the corrector slide measured the requisite level in candles. With the foregoing scaling of illumination and adjustment given by the corrector slide, the illumination of the model terrain afforded by such a projector simulated realistically searchlight illumination. The Argus projectors were mounted on small stands. In front of each projector, there was a solonoid operated flag shutter which could be opened remotely by closing a switch at the experimenter's desk. The switch was a double throw type and could be alternated in its two positions to allow two such projectors to serially illuminate the target, If single searchlight illumination was required, the timing of illumination was controlled in a similar manner. The projectors were enclosed by boxes, except for an opening in front of the lens and shutter, to prevent stray light from altering the general illumination levels. In simulating the proper separation of a searchlight source from the observer, a problem arose in connection with the scaling of our distances. In keeping the searchlight at the same distance from the target as the observer, it was not possible to place it closer than about twenty 8

The University of Michigan * Engineering Research Institute 26991 -F inches to his side (due to the presence of the observing booth). Thus an equivalentseparation of seventy-five yards could be achieved directly but smaller separations could not. Since placing a searchlight siwlation projector in front of the booth displaced it about thirty inches forward of the observer (equivalent in scale factor terms to 90 yards) it was necessary to employ an indirect means of achieving "on beam" and small separations. This was done by setting the projector at right angles to the front of the booth and positioning a front-surfaced mirror at 45 degrees inclination about two feet in front of the objective lens. The mirror reflected the light along the desired forward path. In this ways the searchlight beam could be made to nearly coincide with the observer's line of sight, or to project along other desired paths. Since some light w&s lost due to the mirror, the actual distance from the source to the:arget was redu ed slightly to compensate when this manner of operat'i on was Temployed. Since the searchlight beam originated 44 inches from the floor of the room, the observer's line of sight was a little higher than the seaxchlight beam it$self as would be the case for an observer l.ooking tr&D a position on top of a tantk turret, The appearance of various targets under these different illumination conditions is illustrated photographically in a number of figures to be introduced in a later section. Photomet Throughout the sessions the establishment and control over levels of illumination, and the measurement of values of luminance for selected portcion of terrain and targets $a w complLshded by use of the photoelectrio telephotomter developed in these laboratories by Mr. Benjanin S. Pritchard under sponsorship of the Illuminating Engineering Research Institute, It was necessary to use such a device because direct visual photometry for such low levels of illumination would not be aceurate or even possible in sme instances'The instrument consisted of two units, the photoelectric sensing unit and its associated optics, and the registering meter containing the power source (dry cell batteries). The instrument allowed its user to view through the optical sytemn and see directly that which was to be measured photoelectrically. A reticle indicated the portion of the field which would be effective at the photomultiplier depending on the aperture used as a field stop0 A Wratten 106 filter was effective in correcting the spectral sensitivity charac.terstics of the photomultiplier to approximate the photopice visual curve. It was possi.ble to position the telephotometer at the observer s location, the equivalent of 1000 yards, for example, and register selected parts of a target such as the upper turret of a model tank., The photonau tiplier unit itself could be remaoved from its housing and placed on a terrain area to measure incident light falling onl the area. The complete instrument contained a radium.phosphor internal standard which could be used at any time to repeat a sensitivity setting and allow the meter scale readings to be interpreted in physical unit terms. The instrument could not give information for the worst conditions (lowest light levels, small target areas) but could cover a sufficient range that these conditions can be f airly well understood in any event. Th8 C~ioS.8 EtZ2Y ~l 1XeS0o LB8N.n9

The University of Michigan * Engineering Research Institute 2699-1-F Observers. Three observers served during the experimental sessions. The writer (C.H.), Mr. Carl Semmelroth (C.S.), and Mr. William Dickerman (W oD,). The writer observed in every observation condition reported although, for reasons indicated later, the data obtained for some conditions are not utilized. In any event, the writer has a basis in direct personal knowledge of each observer task posed by the many conditions of the experiments described in the following section. This experience was highly desirable to afford continuity between conditions where different additional observers might be used, and to provide a meaningful basis for relating phenomenal aspects of the observer's tasks to the varied particular situations, Mr.o Seo elroth observed under most conditions, and, in the beginning provided a basis for assessing the effects of practice. Prior to this work he had not had experience as an observer. In addition, he visited during the field tests at Fort Knox, along with the writer. His role included being the experimenter when the wr2iter was observer, and he had complete knowledge at all times of our procedures and objectives. Mr. Dickerman supplemented as observer and assisted the experimentation when not observing. His knowledge of the task was not as complete, but sufficiently so to ensure objectivity during observing. Mr. Dickerman's beginning observation sessions also afforded an indication of the effects of practice since he, too, was initially naive as an observer. The visual capabilities of the observers were checked, For CH,, right eye only, far acuity was 20/18, No measutable central acuity exists for C.H.'s left eye, as explained below. For C.oS,, binocular far, acuity was 20/22. For WoD,, binocular far acuity was 20/18. These measures for CoH. and C.So were with glasses normally worn, and worn during all experimental sessionso W.D, did not require corrective lenses for distant vision, During all sessions binocular vision was employed restricted only by the window of the observer s booth, described earlier. No unusual visual defects are noted for observers C.S. and W oD For the writer, however, the left eye central 20-degree field is lost due to an old scotoma. The peripheral field in the left eye is fairly normal. In the writer's right eye visual field, a small scotoma exists in the lower right quadrant. This scotoma lies about ten degrees from the field center and its presence is phenomenally available to the writer, hence observations can be made with confidence that the scotoma is not interfering. Because of these visual defects, foveal vision is monocular but peripheral vision is binocular for the writer. In spite of such defects, the writer felt capable of obtaining satisfactory data, and similar data obtained from additional observers served as a check in this regard. It should be noted that in the experiments to follow, that our approach was that of classical psychophysics. By this is meant that we, as observers, had complete knowledge of the task and attempted to achieve complete objectivity in making responses. Conditions, to be described shortly, were arranged to provide maximum objectivity, and it is felt that such an objective was closely approached. Some comments relative to the foregoing point il1 be made later when describing the experimental proc~,dures o 10

The University of Michigan ~ Engineering Research Institute 2699-1-F Experimental Procedure: In deciding upon our general experimental procedure, only one previous model simulator study had any direct relevance. This study, by Gordon (Ref. 3) differed in several important respects from the dictates of our earlier conference agreements as to procedure. Gordon, studying the relative visibility of military targets along ground paths, had the observer moved continuously closer to the terrain model until he could distinguish the targets along a line or lines of sight afforded by visible markers at the model edge. This procedure would not be satisfactory since we were to obtain frequency of seeing data, Gordon's targets were sideways rather than head-on to the observer. Illumination was fromn a single source located in varied elevations and azimuths to the observer; the illumination was similar to that from a flare, but fixed in position. Certain of Gordon's findings will be related to our results later, but it should be noted that our experiments called for rather different procedures, and, as well, for a closer relation to specified field conditions. Initially the problem of determining whether an observer reporting the presence of a target was correct led to a procedure shown to be erroneous by exploratory experiments. In view of the paucity of experiments of the present type, the exploratory work will be described and the reasons for its rejection mentioned. In this context, our. final method may be better justified as the most suitable for these studies, Our iritial attempt was restricted to the moonlight condition with uniform terrain surrounding the target. At that time the present large uniform area of the lower left side of the model was divided into two equal sections differing in ground cover, Figure 6 illustrates this earlier arrangement. Five target positions and paired alternate positions were located in the original area. These are shown in Figure 6. The five positions varied in distance from a given position of the observer's booth by fixed amounts. In selecting these target positions, care was taken to select positions such that uniform ground would surround the target for at least twice its own dimensions in any direction, The positions chosen met this criterion, and in fact were the only locations which would allow for the experimenter to rapidly place the target on the terrain for each observation. The experimenter's task was complicated by attachments to the target to be described shortly. During a session, after appropriate time for dark adaptation, the observer was instructed via the intercom to close his eyes. The experimenter then placed the target at one of the five distances according to a randomized schedule. On signal, the observer opened his eyes and tried to detect at which point in the field there was a target. Concurrently with the verbal signal to begin observing, the experimenter closed a switch which started the interval timer. When the observer thought.he detected a target, he depressed the response button on the box at his hand in the booth, This stopped the timer and simultaneously closed a circuit to allow a charged condenser to "fire" a small1 neon lamp briefly. The lamp was mounted under the target on the terrain. This was connected by fine wires to the control circuit. The observer could tell by the flash of the neon lamp whether his response was correct (to a target) or wrong (to some other visual stimulus in the field), 11

The University of Michigan * Engineering Research Institute 2699-1-F The procedure seemed effective and fairly consistent results were obtained for the tank as target over several sessions. Examination of our data showed that a distance could be chosen for the observer from the nearest target position at which he nearly always detected the target. Performance was progressively poorer for the succeedingly more remote target locations Further, in the most remote location, the target was essentially invisible to the observer. Two difficulties led to abandoning this procedure. First, the possible target locations were readily learned in relation to visible terrain features (mainly skyline characteristics). It then was easy to imagine seeing a target in a more remote position when it was not visible in a nearer position, (proven by data from blank trials). More seriously, considering the an:gular size of the target, it should have been visible at greater distances than our data indicated. Accordingly, check experiments were conducted in which the obsesrver-to-target distance was the same as for the earlier sessions with reference to the most remote target position, but the target itself was in the most forward location~ This check showed that the target was now almost perfectly detectable and that its earlier loss of detectability in the more remote position probably was due to its relatively greater proximity to non-uniform skyline (as seen by the observer). It appeared that under these conditions, we had a mixed case in which some targets were in a fairly uniform surround and others essentially in a nonuniform surround. How much separation of the target from discriminable non-uniformities would be required for a target to be considered as in a uniform surround was clearly impossible to predict in advance. Our preliminary work showed that we simply had not provided sufficient uniformity in the immediate background for some of the targets. To correct the foregoing situation, the terrain was modified to provide a greater uniformity of target surround. The grass cover to the right of the original target field was removed and the entire area between the road in the center to the trees orn the left hand edge was made uniform as shown in Figure 7, A new single target position, designated as the "mid-field" position was chosen such that at any intended distance of the observer's booth, the observer saw no obvious lack of uniformity in the surround for distances up to several times the dimensions of the target. Because of this change, the experimenter had to position the target well into an interior portion of the model and it was impossible to do so easily if the neon lamp were attached to the target. Also, some further exploratory observation showed that inconsistent data would result if the distance between observer and target were varied by altering the target position, hence the single target position was dictated. In view of this latter finding, the use of the neon flasher indicating the target position was eliminated and our final experimental procedure was developed. The final procedure adopted was used throughout all the sessions from which data and observations are reported. In this procedure, for a given condition of illumination, the target occupied a single position on the terrain model. For a given series of observations, the observergs booth was located at a fixed distance, and for this series he attempted to 12

The University of Michigan ~ Engineering Research Institute 2699- -F report the presence or absence of the target for each discrete observation. After the series was completed, the booth was moved to a new distance from the target position and another series was run. With this procedure, the problem arose regarding the proportion of trials in a series which should be blank. It seemed evident that one could not arbitrarily say, regarding the type of errors an observer might makle, whether a false alarm (or false positive) or a miss would be more serious under actual field conditions. Hence it seemed desirable to arrange the stimulus series to avoid predisposing the observer in either direction (aside from his inherent non-randomness)o This dictated that during half the trials of a series, the target should be present and half should be blank0 Further, it was felt by each observer that he could dissociate his memory of a previous response better if on any given trial the probability of a target being present was the same as being absent. Accordingly, for each experimental series, the particular target was present for half the trials following a randomized schedule0 The procedure for each observation was similar to the procedure earlier0 To begin, the experimenter instructed the observer via the intercom to close his eyeso The experimenter then went in every instance to the model and physically touched the target position. When he left the model, the target was either present or absent according to his schedule. He then gave the verbal signal "'ready'", over the intercom, at which time the observer opened his eyes and looked to the right of the model (out of the target area) When the observer indicated that he was ready, the experimenter gave the signal, "begin", and started the interval timero The observer then attempted to detect the target if it was present. When the observer reached a decision concerning the presence or absence of the target, he pressed the response button stopping the timer, and indicated his decision by saying "yes" or "no" over the intercom. After recording the response and its latency, the experimenter proceeded in the same manner to arrange the next trial in the series0 During a series, the observer was not told the correctness of his individual responses0 At the end of a series, it was common to record phenomenal observations of interest, and ordinarily the observer was informed of his over-all performance at this timeo In general, for a particular target and condition of illumination, we made an initial guess as to how far away the observer could be and still have some expectation of seeing the target. The observer then ran a series of observations at this distance and if performance was reasonably good, the distance was increased and a second series run. This was continued until the distance was sufficiently great that, for a series, the observer's responses were degraded to chance (being "wrong" as often as "right'")o Our data then show, for a set of varied observational distances, the relative reduction in frequency of correct judgments ("yes"l target present or "no", target absent) in proportion to wrong judgments ("yes", target absent and "no",l target present) Some remarks were made earlier concerning our effort to give completely objective observations. Under these conditions an observer could, of courses keep track of the number of each of his two catagories 13

The University of Michigan ~ Engineering Research Institute 2699-1-F of response and equate their frequency. We did not do so. During a series we might be aware that we had exceeded a fifty percent level of "yes" or "no" responses, but felt definitely that each instance of observation demanded its own unique judgment regardless of preceding responses which might be recalled. Also, we encountered some conditions in which false positives were likely to occur and some others in which they seldom would occur (misses being the rule, instead)o We simply let the situation dictate each response as completely as possible and tried to gain the best possible understanding of it. During a given session, if the observer felt that anything was forestalling the desired objectivity he could (and did in some instances) call off further observations at the time. A final comment relates to the measurement of response latency. During all sessions except those employing searchlight illumination, the observer was allowed as much time as he required for each judgment. For the searchlight conditions, the observer was not given unrestricted time. Instead, each series featured a predetermined fixed time during which the searchlight or searchlights illuminated the target position. Reasons for this procedural change are discussed in the context of the specific experiments

The University of Michigan ~ Engineering Research Institute 2699-1-F III.. EXPERIMENTAL RESULTS The experimental sessions may be most conveniently organized for reporting in terms of, first, terrain background, and, second, mode of illumination. With this organization, comparisons may be made readily among the results for the different target types and from the different observerso In presenting the results from our experiments, the main data are shown in tabular form. For each condition, in the order introduced in the text, a table shows the frequency of response data and, except for the searchlight conditions, the time for response (latency) data. In analysing these data for discussion, a statistic is derived (also shown in each data table) which is used for illustrating the change in target detectability under the particular experimental conditions employed. Since the statistic referred to, above, was developed for the purpose of these experiments, some discussion must be given to its rationale before proceed`-og.g We may begin by considering the results of one group of sessions by way of illustration, These sessions, shown in Table I A are for the M-48 tank in a uniform background under starlight only. The responses fall into four catagories: (1) "yes"g target present; (2) "no", target absent; (3) )"yesl" target absent; (4) "no", target present. The first two catagories are correct responses; the second two are wrong. It may be seen from the table that at the scaled distance equivalent to 135 yards, the observer (C oS o) made no incorrect responses. As he was moved successively further from the target, he made fewer correct and more incorrect responses, until at 225 yards, he made about as many wrong as right responses. The changes noted in response frequencies for each catagory are shown in Figure 9 with the open circle points. Each open circle is a proportion, shown on the left ordinate for the catagory of responseO The change with increasing distance may be noted easily. In considering how a single value may be used to represent these changes in response, it might at first be thought that if the proportion of detections ("yes", target present responses) were corrected for the proportion of false positives ("yes", target absent responses) a usable quAtl'ity would be obtained, Such a procedure would be essentially the conventional form of data treatment used in a "yes - no" psychophysical experiment. However, there is now ample evidence (Ref. 4, 5, 6) that critical problems surround this form of analysis due to questions concerning the observer s criterion level for responding affirmatively. Without examining this problem in detail, it may be noted that our experiments allowed for a high proportion of false positive responses, and were done undez condition's5hat criterion shifts were almost inevitable. In connection with the latter point, comments will be added later concerning the physical conditions which literally forced such changes upon the observers. 15

The University of Michigan * Engineering Research Institute 2699-1-F In general, it was felt that a highly ambiguous measure would result for our data if the detection proportion corrected for false positive responses was used. Alternatively, it was felt that our data might more meaningfully be interpreted in some other terms where their characteristic features entered a single representative quantity in an unambiguous manner. These considerations led us to think in information terms, where the assumption is rmade that we can take the technical sense of information and consider that it has a reasonable counterpart in the conventional sense. Granted this assumption, we can treat our data for the specific information content of the set of responses for a given condition and arrive at a measure of stimulus information to be gained from knowing the responses. The basic derivation, due to Dr. Wilfred M. Kincaid of these laboratories, leads to the following equationo 4 = 1 + 3.32 [P(Y) log P(Y) - P(N) log P(N) + P(SY) log P( Y)+ P(SN) log P(SN) + P(S'Y) log P(S'Y) + P(S'N) log P(S'N In which M = stimulus information gained from knowing the response (the representative quantity to be used) P(Y) = Proportion of "yes" responses to total responses P(N) = Proportion of "no" responses to total responses P(SY) = Proportion of "yes" responses made when a target was present, to total responses P(SN) = Proportion of "no" responses made when a target was present, to total responses P(S'Y) = Proportion of "yes" responses made when a target was absent, to total responses P(S'N) = Proportion of "no" responses made when a target was absent, to total responses 3.32 = Constant to convert to system of logarithms base 2 Under the terms of the foregoing equation, if the observer always responds correctly, maximum information is gained from his responses concerning the presence or absence of the targeto Since the target was presented during half of the observation trials, the maximum value for a trial would be one bit as an information quantity. It may be shown, also, that if the observer always responds "yes" (or 1"no") or~, if he responds correctly and incorrectly equally often, M goes to zero indicating that no information is obtained concerning the stimulus knowing only the observer's responses. It is the case, too? that if the observer is always wrong maximum information results which should not be surprising (although this did not occur during our experiments). Some further features of the statistic and additional comments are given in Appendix B,. For purposes of further description of results, the information quantity, M, will be used as defined in the foregoing equation. Returning to the data shown in Figure 9, the value of M4, shown on the right hand ordinate, is plotted for the set of responses at each observing distance. It may be seen that at the nearest distance, _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 16 __

The University of Michigan * Engineering Research Institute 2699-1-F 135 yards, the observer gives maximum information. As he was placed successively further from the target, the value of M falls until at about 200 yards it reaches a value so low as to be no information at all. In the material that follows, the information quantity M will be used to represent observers' levels of performance, and by direct implication, target visibility, as the dependent variable in graphic representation in a manner similar to the foregoing example, Results from the specific sets of experiments follow, These results deal with detection only, and information concerning target identification is given in a later section. Photographic illustrationrfor several different experimental conditions are contained in this report. Not every condition could be photographed, Reference to these illustrations will be made at the appropriate points~ 17

The University of Michigan * Engineering Research Institute 2699 1-F A, Experiments With Uniform Terrain Background (1). Starlight Illumination An initial set of experiments studied the targets under this condition with three observers. However, a serious error in procedure was brought to light and these data were discarded, The condition was repeated with Co.S as observer and the results are expressed in Tables I A, B and II A, B for the M-48 tank and M-59 APC targets. Tables I A and II A show the frequency-of-response data, and Tables I B and II B show latencies for the respective conditions, In Tables I A and II A, the figure in parenthesis following each frequency of response figure is the proportion for that frequency to the total number of observations, Similar tables are given for the remaining experimental conditions, Figure 10 shows graphically the results for this condition, The illumination was so low that the anti-tank gun target (referred to as ATG hereinafter) could not be seen from the observer's booth at the nearest practical disposition on the terrain which means that its visibility distance would be less than 120 yards, From Figure 19, it may be seen that the tank and APC show a very similar pattern of loss of visibility with increasing distance, Both remain fairly visible to about 190 yards and then rapidly are lost to the observer. It should be noted that in the terrain position used, the targets were about ten inches (equivalent of 30 yards) lower than the observer, When they* were nearer, he responded to a disproportionately larger target since he could see more over their top. The targets could be seen only in peripheral vision and appeared as indistinct black lumps against the faintly visible ground. The presence of any non-uniformities (as will be noted later) made the observers task completely impossible. It is interesting tosnote that under this worst condition of visibility but with a uniform surround for the target, its visibility remains fairly good as far as it does. Discussion of the point will be given in a later section, but iC may be noted, now, that the target here acts much as in more basic psychophysical experiments with reference to angular size of the target and contrasts present, (2), Moonlight Illumination Tables III A, B, IV A, B, and V A, B show the results for the tank, APC, and ATG targets respectively, under moonlight illumination. Figures 11, 12, and 13 show these results in graphic form. Figures 21 and 22 show the vehicle targets for this condition. It may be noted that under moonlight, the tank does not appear to be highly visible to observers at any distance shown. The nearest distances employed in this instance were the first found, informally, which did not yield high or perfect detection levels, Hence at 700 yards, for C.Ho and C.S.o M would have been 1.0l This was true for WoDo at about 600 yards. Beyond these distances, performance was less than perfect as shown in the figure, It was noted for this target and the others as well, that it was detected entirely in peripheral vision as a dark target, Essentially~,it was a black blob which, at increasing distances, was very hard to pin down and was never seen with 18

The University of Michigan * Engineering Research Institute 2699- 1F high confidence. In the later discussion section, more consideration will be given to the phenomenal aspects of the target. The APC target appears generally to be a little more visible than the Mw48 tank under these conditionso This finding is reversed for later searchlight conditions. Greater observer variation is seen alsc for this target. It may be noted that observers CoHo and CoS show a reversal in the trend of their data. This corresponded to a real reversal of a phenomenal nature. For both observers the target at first became more difficult to see (with increasing distance) and then, at a particular distance, became anomolously visible. With further increases in distance, the target became impossible to see- A similar phenomenon was noted for the ATG target although it is not as well reflected in the data, By way of accounting for such reversal, it may be noted that while the target surround was relatively uniform, it was not absolutely so. Thusa at a sufficiently close distance, the observer saw some non-uniformities in his field of view. By virtue of the relation of target size (and shape, probably) these residual non-uniformities were not easily confused with the target, However, as the viewing distance was increased, the target became less distinct and greater confusion was possible. However, since the non-uniformities in the field were minimal, it is probably that with even greater observing distances, they went below threshold leaving the target somewhat more visible because it was then in a virtually uniform background in comparison with the earlier situation, This would give rise to the anomalous result noted and attest to the complexity of the perceptual task. The question as to the greater visibility of the APC compared with the tank is reserved for later discussion. The ATG target, as shown in Figure 13, is considerably less visible than either of the vehicles used, One observer, WoDo continued to report the presence or absence of this target with some accuracy at a surprisingly great distance, but in general from 400 to 500 yards, it was essentially undetectable. As with the other targets, this was detected only peripherally and was only a dark target. (3). Searchlight Illumination Under these conditions, the targets were illuminated by a single searchlight either (a) along the observer s line of sight, or (b) displaced from the observer the equivalent of 75 yards. The searchlight duration for a given observation was 1, 2, 4, 8, or 15 seconds. Also, two searchlights were used each displaced the equivalent of 75 yards on the right and left of the observer respectively. The two lights alternately illuminated the target at a 1-second flicker rate and provided continuous illumination for 2, 4, 8 or 16 seconds. Figures 23 and 24 illustrate the appearance of the APC and the anti-tank gun and crew target illuminated by a single searchlight displaced to the right of the observer. Under all these conditions the tank and the APC targets were always detected even at 1500 yards range. For this reasson, extensive data were not obtained. The ATGC target, beinag smaller and more irregular in shape, Was never detected at 1500 yardso ___________ __________ ________ _ 19

The University of Michigan ~ Engineering Research Institute 2699-1-F Considering the ATG target further, when the observer was moved to 1200 yards from the target and the searchlight was displaced 75 yards to his right, the target was detected fairly easily for searchlight durations as short as 4 seconds. With further reduction of the duration of illumination, performance became poorer as shown in Table VI. Two features of the task were noted. First, the target itself was not detected, but instead its shadow, thrown to the left was seen. This was in distinction to the fairly visible outline of the vehicle targets in addition to their shadows. Secondly, the anti-tank gun and crew probably could be seen at somewhat greater distances if sufficient time were allowed. The effect of reduced illumination time was to given the observer too little time to seek out the target. There were few false positive responses. If the observer had time tc find the target, the response was definite; if not, there was nothing else which could be confused with it, One further set of observations was made with the ATG target under these general conditions, For these, the searchlight was along the observer's line of sight. At 1200 yards, the target could not be seen at any duration of the searchlightS When the observer was moved closer to the targets it suddenly became visible at between 900 and 1000 yards, This appeared to occur because at this distance, the shadow of the target, previously concealed behind it, became visible over the top of the target. In fact, the shadow outlined the target very plainly making it highly identifiable in comparison to when the shadow was visible only as thrown to the side by the earlier searchlight condition, The effects of searchlight flicker were studied informally in the foregoing conditions. It appeared that the targets were so visible under all the searchlight configurations that it would be more profitable to reserve extensive study for these targets in non-uniform surroundings. Under such modified conditions, usual field conditions would be more nearly approximated and the results should have more applicability~ It was noted, however, that with flicker the observer gained an advantage in that the target shadow "wig-waged" and became a little more noticeable. The question remained at this point as to whether, with other objects in the field which would become more notieeable, confusion might occur with flicker. Information relating to this point is given in a later section,

The University of Michigan ~ Engineering Research Institute 2699- -F B. Experiments With Non-Uniform Terrain Background For these experiments, the terrain was modified to add bushes and trees in the areas shown in Figure 8, The observer looked for the target in a fairly clear alleyway of vegetation. Three target positions were used, also as shown in Figure 8. The different target positions permitted an evaluation of differing degrees of non-uniformity of imediate surround. The target in the "mid-field" position, used in the moonlight and searchlight experiments just described, was still seen with a fairly uniform surround. Brush and trees were closer for the "upper field" position. When in the "tree position", the target was seen againstAclose background of vegetation. (1). Moonlight Illumination Results for sessions with the M-48 tank target in the mid-field position are given in Table VII A, B and in Figure 14 for observer C.H. The appearance of the target is seen in Figure 25. This figure may be compared with Figure 21 to note the changes introduced by the terrain additions, This condition is completely similar to the earlier condition except for the non-uniformities added to the terrain. For this reason, it permits a comparison of results showing the effect of the non-uniformities added. The present results show that the target is fairly visible up to 500 to 600 yards in the moonlight alone. Beyond this, it loses detectability and by 700 yards, is essentially not detected. Referring to Figure 11, the same target and illumination with a uniform terrain, it may be seen that at even 900 yards, the observers were still responding with fairly good accuracy. (2). Searchlight Illumination A much more thorough exploration of the effects of searchlight illumination was carried out for this condition of terrain than was done earlier. Because there are several variables which can be manipulated in this situation, the results can be organized in a number of different ways for presentation. For this reason the several remaining tables will be cited first, and data drawn from them for more meaningful graphic presentationo Table VIII gives such data for observer C.H. and COS., for the M-48 tank in the upper field position and several distances from the observer. The various searchlight configurations are shown in the left hand column of the table. SL-C means that the searchlight was projected along the observer's line of sight. If the searchlight was displaced from the observer, this is shown as to distance and direction of displacement. The flicker conditions all used the same displacement of two lights. One was the equivalent of 75 yards to the left and the other an equal equivalent distance to the right. For the flicker sessions, the -e'"t figure is the duration of a single light flash and the scoradthe total continuous time of illumsination. Table IX shows similarr i the tank target in the tree position, for obsrerver CH, _ _ _ _ _ _ _ _ _ _ _ __ ~ 21

The University of Michigan * Engineering Research Institute 2699-1-F Table X shows results for the APC target in the upper field position in a like manner for observer CoH. and C.S., and Table XI gives this information for the target in the tree position. -Data for the remaining target, the anti-tank gun and crew are shown in Table XII for observer C.H. Illustrations of the appearance of the vehicle targets illuminated by a single searchlight displaced to the observer's right are shown in Figure 26 - 29. Figures 26 and 27 show the tank in the upper field position with the' searchlight displaced 75 and 150 yards respectively. Figures 28 and 29 are similar illustrations for the APC target. Smaller displacements for the searchlight cannot be photographed ewdily since the camera must be closer to the target than the searchlight hence interferes with the illumination. Illustrations are not shown for targets in the tree position since they are practically impossible to distinguish in a photograph at this location. Combining data from Tables VIII and IX for the most meaningful comparison leads to Figures 15, 16, and 170 Figure 15 shows the change in visibility for the tank with increased distance of the observer from the target. This comparison is for the condition in which the target is in the tree position, illuminated by a single searchlight along the observer's line of sight for a duration of 4 seconds per observation. It may be seen that under these conditions, performance is fairly accurate at 1000 yards but gets progressively poorer with increased distance. Since the foregoing information is all for the searchlight on the observer's line of sight, the effect of varying its displacement from the observer is to be shown next. We selected the single observing distance of 1350 yards since performance was relatively poor (P =.193))with the expectation that displacing the searchlight from the observer would result in some degree of improvement. Figure 16 shows our results arranged to show such effects, Two durations were employed, The first we emioyed was 4 seconds, As anticipated, performance improved being as good at 40 yards as it subsequently was at 75 yards displacement. With greater displacement there was some loss in performance which will be commented on later. It is possible to show the effect of flicker, here, also. The single point labeled 1-sec-F, 4-sec-total is shown for this purpose and is noticeable lower than the 4 second continuous illumination curve at the 75 yards displacement. The second curve labeled 8-sec duration in Figure 16, is not in error although it at first might appear so, After results were at hand for the 4-second condition, the 8-second condition was employed to assess such further improvement as could result from additional time to search for the target. Instead of improvement, the opposite occurred. This result was sufficiently startling that it was checked with additional observations, as was done for c..ertain of the points on the 4-second curve, The finding, to be discussed more fully later on, held up in that additional observation time simply degraded performance for this target under these conditions., 22

The University of Michigan * Engineering Research Institute 2699-1-F Considering that the foregoing results might be due to unique circumstances in which the tree background became more confusing with additional observing time, the effect of both duration of the searchlight and the background were assessed with the target in the upper field position as shown in the next figure, 17. In this figure, the target distance is fixed at 1512 yards. It was necessary to use the extreme distance because at nearer distances, performance for the more open upper-field target position would be practically perfect even at the shortest searchlight duration. In this analysis~ curves are shown for each target position and performance is plotted as a function of searchlight duration. For both target positionsj increasing the duration of the searchlight illumination only leads to improvement of the observer's performance. The difference in visibility for the target for the two locations in the terrain is quite marked. Also, it may be seen that the greatest gain in performance is for the target in the more open position and occurs mainly within 10 seconds. Further comment will be made later relative to the latter point. The dashed line curve shown in Figure 17 allows a comparison of searchlight flicker at this point, also. The flicker data, also, are for the target in the upper field position. At the total illumination times of 4 and 8 seconds, performance was definitely better than for equal continuous illumination durations, The shadow of the target quickly "wig-waged" the observer to a detection from this relatively open location in the terrain. With 15 seconds total flicker at the 1-second rate, the task became phenominally more difficult as reflected at the right of this curve. Considering next information for the M-59 APC target, data are shown from Tables X and XI in Figure 18, 19 and 20, In Figure 18, the visibility of the APC for two conditions of searchlight displacement (SL-C and SL-75-R) is shown as a function of distance of the observer from the target. It may be seen that with the searchlight projected on the observer's line' of sight, the target is detected only poorly as close as 1000 yards compared with the instance where the searchlight is displaceg75 yards to the right of the observer, For the latter situation, performance is very good to nearly 1200 yards and then drops very rapidly to become as poor as with the first configuration of the searchlight, In general, the APC is less visible than the M-48. tank and reasons for this will be given in a later discussion. It was possible to graph two additional points for flicker conditions in this figure, also. With 1-second flicker, performance is noticeably poorer than with continuous illumination for the same total time, When the flicker rate is decreased to two seconds, performance improves. From Table XI, it may be seen that if the 2-second flicker is continued for a total of 8 seconds, performance becomes perfect. The effect of differing searchlight durations are shown in Figure 19 for the APC target. Data are shown for two observers. For this target position, increased duration resulted in improved detection. Two sets of flicker observations are shown also. Again, for a target in the fairly open upper field position, some improvement is noted for f'icker compared with the condition of continuous illumination. 23

The University of Michigan * Engineering Research Institute 2699-1-F Figure 20 shows the effect of searchlight displacement for the APC target, In Figure 19 it already has been shown that a substantial improvement in visibility results for this target when the searchlight is displaced 75 yards to the right of the observer, compared with the SL-C arrangement, The present Figure indicates that the improvement results primarily after displacementsof about 60 yards have occurred, With further displacement of the searchlight, as with the tank target, some loss of visibility appears to take place. Also, as with the tank target, the analomous result that doubling the duration of the searchlight illumination produces a reduction in performance level for the observer is noted)although to a lesser extent, Final results concern the anti-tank gun and crew target. Data for the limited study given this target are shown in Table XII. From this table it may be seen that at approximately 900 yards the visibility of this target in the fairly open upper field position ranges from good to poor depending on the searchlight duration and displacement. The best results were obtained from the flicker condition shown, However, increasing the observers distance only an additional 200 yards made the target very difficult to detect. When the target was placed in the tree position, it became impossible to detect at all even at much closer distances. Since the target was fairly visible at as great as a range 1200 yards under searchlight illumination on the earlier uniform terrain, the effect of concealment for this type of target by vegetation is fairly well demonstratedo In the case of this type of target, if there are nearby bushes or trees, confusion results; viewed against bushes or trees, such a target readily blends into the background as neither the tank or APC can do. The final comments in this section concern the response latency data shown in certain of the preceding tables. In order to keep this report within some limit, no detailed analysis will be presented of these data although they will be considered further in the next section. It may be seen on inspection that the average times taken by observers was fairly consistent depending on the catagory of response and the difficulty of the immediate observing task. The shortest response times are seen in general for'"yes", target present' responses while the false positive, or'"yes", target absent' responses generally take as long as a correct or incorrect "no" response. All responses tend to increase in latency as the observer's task became more difficult due to increased observing distance, with the correct positive responses increasing the most. Over a variety of conditions it may be noted that the range of average latencies was from as short between 5 and 10 seconds to as long as 50 to 60 seconds. The three observers differed in latencies typical of their responses. 24

The University of Michigan * Engineering Research Institute 2699-1-F IV* DISCUSSION OF RESULTS First to be considered are the results from the starlight and moonlight conditions. It was noted earlier that under these conditions targets were phenomenally dark targets (relative to their background). An observer was able to detect them only by using peripheral vision. An observer's ability to do this seemed to depend upon some kind of scanning technique in which a distinguishable terrain feature (such as a particular patch of trees in the extreme background) was used for relative orientation. When attempting to judge the presence or absence of the target, the observer scanned in the direction of the orientation feature of the terrain but attended to the peripherally aroused sensations in the area where the target should be. Scanning was necessary because of the generally indistinct appearance of terrain features even under moonlight. It should be added that since between the observer and the model there was essentially a void, nearby terrain featurev present in the field were not available to aid orientation in our simulator studies, When observing even a fairly close target under these conditions, it could be held in phenomenal regard only a short time. when the observer's distance was increased, the target tended to appear less frequently and disappear more rapidlyo The limit of a given target's visibility was characterized by such a fleeting appearance that the observer generally lacked confidence in his judgments entirely. All observers were impressed by the "noise" inherent under these conditions. Although the terrain appeared uniform in the target area, it was very diffuse and the peripheral phenomena sensed often were as convincing when the target was absent as when it was present. The task, generally, was very difficult. It was noted earlier that our exploratory experiments of the tanlk target in moonlight led us to believe that we had failed to achieve the maximum distances for detection for the uniform terrain condition. Following, we developed the altered procedure reported. In connection with the later findings it was stated that the target, under the starlight and moonlight conditions appeared to behave very much as targets in a more basic psychophysical study for uniform targets against a uniform background. To supplement this point it may be noted, from photometric information, that a reasonable value for the tank contrast with its background for starlight and moonlight would be about -.50. Considering the background to have a luminance of 10-5 foot lamberts and given this target contrasts the minimum visual angle for the target to be detectable at the 50 percent level of probability can be estimated crudely from visual detection data (Ref. 7). Under these conditions, about 40 minutes of arc represents the minimum detectable angular size of a target. This would correspond to the APC or tank about equally well at about 300 yards from the observer. Under these conditions our targets remained visible to slightly over 200 yards and were still detected very occasionally at 300. The complexity of having no discrete fixation point for the observer and having, actually, a relatively non-uniform field of view (compared with the more pure psychophysical case) make for losses of detectability of about the order to account for our results. 25

The University of Michigan ~ Engineering Research Institute 2699-1-F A similar analysis made for moonlight conditions shows that the threshold detection angle would be about 7 minutes of arc. This would correspond to the width of one of our vehicle targets at about 1850 yards rangeo Again, residual non-uniformities in the field of view, lack of discrete and consistently optimal orientation, and other complexities can be cited to account for not reaching such an extreme visibility distance under this condition of our study. However, under the above mentioned conditions we have shown visibility distances approaching limits of observer sensitivity. These results exceed the performance conventionally expected or seen, given only starlight or moonlight illumination. While some qualifications will be made shortly, our results suggest that certain observer capabilities along these lines might be worthy training objectiveso If observers could take better advantage of existing natural illumination, possible targets may be located prior to more revealing active observation is initiated using searchlights Probably at best, under starlight, unaided vision is not particularly helpful. However, given added auditory and motion cues under these conditions, possible targets can be located and supplementary optical aids employed for the advantage they might give. Under these conditions, when a target seems probable, even a brief duration of searchlight illumination would make the detection and identification highly certain~. Under moonlight illumination a greater range of visibility was seen. The remarks of the preceding paragraph are considerably more pertinelnt here o The observer can scan a fairly large target area. Since he is not revealing his position by doing so, an unlimited time can be provided0 (Actually, for a given area detection may occur in a fairly short time so observing techniques might well involve discrete observations in specified areas). If an observer has had the advantage of previous reconnaissance under better illumination, the moving of a target into the. area may well be seen by him within the limits we have notedo Optical aids then may be adequate for complete identification or at. least a more certain detection. At all distances that moon illumination is adequate for some degree of detection, the increase of detectability and identifiability with added searchlight illumination is very marked (given an optimal searchlight configuration, of course). Two additional questions concern our findings under these conditions of illumination. First, to what extent are they representative of what may be expected in the field. Second, what are the roles of training and motivation to achieve such resultso Following discussion of these questions, attention will be given to the factor of identification and to the degrading effects of increased field non-uniformities. With reference to the first question, above, the visit of the writer and observer C.oS for the field studies at Fort Knox gives some indication of an answer o uring the first run of the evening of June 27, the moon was nearly full and oriented with respect to the target area a26

The University of Michigan * Engineering Research Institute salmot exactly as in OT tilatoT stuies.o es =lng this rur it was, possible to look for either $a M-48 tud or M-59 APC target at distances of 500, 700 1000 or 1175 yards after they movhd Lnto position but before the searchlight was terned ono Dark adiptataog cou~ld not be optimal because of the presenca of the sechslc hgt m well as cther lluMIn nrati on tn the observing stEre&o Also, there was a mmldly danse ground fog visible in the target areao Hoevver: both the wri te ae.d Co.S could detect the presence of the tank or the AP at 500 ad 700 ywads with the unaided eye. At 500 yagrds. detection was (certinan; at 700 it was poorFer but still occurred. The preceding figures' compal wel with out model study findings for the nonC-ri form t~e=rl&o {ad the goronn fog been absent., someviat greater visibilit y wol ld iav rtsulted althoag with the highly nonUnforma terrdain condIfAtns probably nost ch improvemnt would be seen. It may be added tht theese tagsets *ean detectedo were identifiable using seven power binoculrso Fimaiyo it should be added that at the field test and in the model study, the Moons aguElwr location relative to the tsaget po$ ston and observer was optimal for target detection as may be inferred from Cardon "s $tuidy' itead prsvizuis y (Bef 0 3)0 The roles of motivation and trTV-nin.g can only be discussed in a speculative manner.,Howe vier Oav snp&,vwei Ce5. d ring the course of the studies give rtse to defint a opinonmg coic2trning these variables. In the first pla e$ it si b been shown that txned o8servers in p sychophysical experimsnts cXa perf mr better thai.s naive Rf o 8). Irn the more omplex field situations the differences between trained aud untrained performance may well be greater. ruing ourt eeare e.~periments, the writer (already a "trained" observzer in the genirtl sens) probably achieved peak performance during a few exploratory sessions. H owever, observer C oS o. and subsequently Wo DP began observin g seassions relatively naive to the nature of the tasko For four cona ecutive two-hoz' sessions, observer CoSo could not perform better than chance wh en he Owas more than about 650 yards range. Nearly up to this point, his respon8es were perfectly accurate, (One should note this range in connec(tionon rith csmmente to follow concerning identific ction of targets later) o Following these Sessions, two adittional sessions resulted in a fairly rapid impkovemenxt (within each session) in which the target could be dtected t successiveLy grea'ter distances,: When C1 o.osB performance beca me stable9, ie wa ls ta o give some) etections at nearly 1200 yards range. His account of the canhageE in aates the learning of appropriate scanning and fL;xation tschniques gnd;i, moTa importantly, to t"te ase out" the phenomenally vagute anrd intag ible target from the unfamiliar visual display. Recall ng hatE the obs:erver was not info rmd of the accuracy of his responses during a $serE s of observation s it may be inferred that more rapid Lmprovement would result with imediate reinforcement of responses, Fairly exxtensive improvement was seen, howevero Similar improvement occurred f or observer Wo.D. and in addition, he did not improve sbstantially untit he became jii.ng to raepon~d ~firmatavely even though the rtsponse might be a faesl po$itlveo Following this shift inL motivation (or set) his implovemant ws nearly as g eat ~ for C.$o It may be noted Tin e f ti fiaB dtte rfige h eat g oo roC b eroe thobsve in addition t o the werWi er, L e f nt>al perf5oraF.nc. as gos;o bset@e th the itC ersBo

The University of Michigan ~ Engineering Research Institute The postbilit.of u;ifrg a, model'i J to for.raining purposes.i evident and shtOzzld be cn$l I8Ted1o There is, of course, precedent in various instances in amiit:%ary training progr,0s. Once constructed9 a morel. ilatr of fers a. relatively inexpensive and readily available means of preseatnng information and promoting acquaintance with field conditionso Conditions can be highly controlled0 Demonstrations of the capabilities of an observer as well as better exposition of techniques employed are readily possible~ The step is surprisingly short from a well simulated field situatton to actual field conditions0 Usi.ng a model simalator for training purposes requires evaluation, howeveri in its own right o Face validity is high and, perhaps, has been misleading in the past0 Earlier progrsam have not shown clearly that any pa~.t'd c'i ar advantage was gained by observers so +-rained. Such failure in the paat has been due, in the writer~s opinion9 to first, not having specific and measurable crLersa for perman E a as objectives during train"i.ng9 anid, secondly, not having similaT criteria for subsequent evaluationo Generally it has been agreed that such devices are interesting and motivating but the actual effects on field performane have been difficult if not impossible to evaluate. Our study, as an investigation of observer capabilities9 is quite specific as to measurable aspects of performance0 (In fact9 the pattern of a training study was reported a little earlier0 Having at hand performance data. one ained obsever t was p ible to continue practice on the part of new observers until they had reached similar levels of performancea) Possibilities for training use certainly merit considerable further investigat'ion In our original outline It was noted that identification data were soght.o The earliest studies showed clearly that the vehicle targets would alw ays be identifted as such withouat confusion with the anti-tank gun and crew target0 This was because of the difference in visibility distance~s involved and because the observer was generally aware of the range to the target area0 For the two vehicles under moonlight and uniform terrain conditions, one set of observations were taken with the writer as observer to estimate the maximum range at which eac h vehicle could be identified. The APC could be distinguished as such out to about 600 yards and the tank a little furthero Actuallyv, in both instances the' targets were only vaguely identifiable at eve~ shorter ranges, due probably because they were viewed in peripheral vision (about 5 degrees from central) Within the ranges noted, the tank had a vague humped appearane whereas the APC was sqauat and more spread out in appearance Beyond the ranges noted the targets remained readily detectable, as shovm earlier9 but quite indistinguishable o Data presented in Figure 14 showed that the tank visibility under moonlight in the non-uniform terrain condition was very poor beyond about 600 yards although it had been considerable better for the uniform terrain condition. This range fordete tion in non-uniform terrain appears to cortesponi to the range for id entificatjion noted above0 It was confirmed in the rter 0s experience that when the k7talram was non-unniform the task became essentially an identl fication t.aa k o Any number of equally detectable via ul. sItimuli were prtesent along Awth the target and the character28

The University of Michigan ~ Engineering Research Institute 2699-1-F istic shape of the target had to be perceived to some extent in order to achieve any accuracy of responding. When the nature of the observing task is considered for the searchlight conditions), the role of target identifiability becomes even more pronounced. Probably for most conditions when the visibility of the target used was less than perfect, factors of confusion with other visible stimuli were basic to the results. This interpretation is plausible when the detectability of the targets is assessed as was done earlier for the conditions of natural illumination. In Figures 30 and 31 several figures show the principal photometric values which can be assigned portions of the targets under typical searchlight illumination. Figure 31 shows these values for the APC target illuminated with the searchlight at 1000 yards, displaced 75 yards right of the observer. Figure 23 shows this target much as the observer saw it. Two aspects of the target are evident as detection features. The horizontal front panel and treads form a roughly rectangular bright target against a somewhat darker background. The sloping front panel, not reflecting as much light to the observer, forms a roughly equal sized rectangular dark target with the background. In addition to the target, its shadow may be seen and this forms a high contrast dark target with the remaining background. Considering only the rectangular areas noted, the lighter has a contrast of about.70 with the general surroundings and should be detectable to an observer at about 3000 yards range on the same line of sight. The dark target area has a similar contrast of about -o40 and would be detectable to a distance of about 2100 yards. The shadow obviously would be more detectable in these terms than the sloping front. Such an analysis is more difficult for the tank and will not be attempted. Reference to Figures 26 and 31 shows similar but less well defined areas of high contrast with the background and within the target as well. Consequently, such analysis if successful, should show this target to be detectable well beyond the maximum range of interest. Of course, if both the observer and the searchlight are moved further from the target the background level is reduced. To some extent contrasts may be changed alsoo However, this reduction in level of illumination would still leave the vehicle targets highly detectable, especially if their shadows are visible as well.o If the targets are detectable, in principal, at ranges where we have already shown their visibility to be quite poor, then the loss of visibility must be accounted for. Our data shown for both the tank and the APC in the successive mid-field and tree positions support the contention that as the background becomes more non-uniform visibility suffers. To the observer, the task becomes an identification task because not only can the targets be detected but so can everything else in the immediate surround. The problem becomes that of distinguishing the target from the other objects seeno For the tree position this is very difficult indeed. These terrain conditions, incidentally, were very similar to those of the field study at Fort Knox; our informal observations in the field with searchlights compare closely to our results from the corresponding conditions of the model study. 29

The University of Michigan ~ Engineering Research Institute 2699-1- F Two results for searchlight conditions were noted which may be explained more adequately now. First, it was noted that the tank target was more visible than the APC under searchlight illuminationo'This was particularly true for the targets in tle tree position. Reference again to Figure 31 shows that the luminanc A'ee background was somewhat lower than the terrain ground itselfo In this location, the sloping front of the APC goes to lower contrast with the background. Also, the vehicle. shadow may be lost due to additional ground shadows. This leaves only the smaller area of the vertical front panel and treads to be detected0 When one success fully locates the front panel under these conditions, it is fairly visible0 However, its dimensions are smaller in proportion to distinguishable aspects of the surroundings and much harder to find0 Visibility was lost accordingly o. The tank, on the other hand, has its internal contrast features distributed to form a single pattern0 Parts of the tank might become low in contrast with the tree background but the over-all pattern would not be changedo Also, the tank being a taller vehicle tends to overlay its shadow on the tree background more than the APC. Consequently, it seems reasonable that the tank should be more visible under these conditions than the APCo In the variation of searchlight duration described earlier, it was noted that for the most non-uniform terrain condition, there appeared to be an optimal observing time0 Increased time beyond this amount actually appeared to handicap the observer. The writer's experience tells what happened. When the searchlight came on, the target area was quickly scannedo The most likely thing to look like a target was, of course, the target itself Hence, if enough time was afforded to scan the target area, the first impression was likely to be correct (regarding the presence or absence of the target)o If more than this time were allowed, there were many parts of the target area which could be imagined as a target. The tree position target location was a veritable "ink blot" situation for targets if one studied it for any length of time. Hence, given the additional time, the observer formed a judgment and then tended to lose confidence in it before being required to state ito This led to definitely degraded performance. Probably the optimal duration would vary among observers and with the specific situation. If knowledge of the probable target location is good, some fairly short duration of the searchlight flash may not only be adequate but best for the tasko In our studies knowledge of target location was high. We always found, however, that a searchlight 1-second flash was too short to scan within the area illuminated, and 2 seconds quite marginal in the same respect. A flash of 4 seconds appeared optimal, as noted, for targets located in the tree position. For the upper field condition, greater visibility ranges were obtained and it was evident that 4 seconds of searchlight illumination did not lead to as confident judgments as when 8 seconds illumination were provided. Doubling the searchlight duration again led in no case to any further substantial improvement on the tasks.o It is the writer9s opinionra, based on these studies and observation at the field tests that from 8 to 10 seconds is adequate for any attempt by a solitary observer to judge the presence or absence of a target in the beam of the searchlight if the probable target locations are reasonably easy to identify. 30

The University of Michigan ~ Engineering Research Institute 2699-1-F The effect of flker probably merits much further studyo Some conclusions can be drawn from our experiments, however For targets in the open9 flicker rather definitely improves detectability due to the "wigwag" effect o In part this may be due to the fact tha- the shadow of the target may be detected more readily than the target itself o Many times it is probable that only the shadoi is seeno Unrder flicker conditions, the shadow appears alternately at one boundary then the other boundary of the target. Under these conditions9 two borders of the target are seen in succession, and the target is more readily perceived as such the shadow, moving with flicker ifs nof misperceived as a substantial object0 With non-uniform terrain9t however, the same f actors may make every bush and large rock equally improved in visibility an lead to poorer observer performance instead. Our experience with flicker showed that it required some getting used too Initially, when one searchlight came on9 the observer would scan forward along the ground path of the light beam0 As this went off and the other searchlight came on9 he turned his direction of sight to scanning the second light path0 This resulted in considerable overshooting and retracing which was confusing as well as fatiguingo With further experience9 however9 observers found that they simply directed their lIne of sight towards the terrain and within the fiet two alternatiorsa 9 would lock-'n on the common area of illumination0 In this regard9 the lsecond flicker rate was always uncomfortably fast whereas- the 2$second rate was quite acceptableo As noted earlier, when flicker was continued very long (15 seconds) the observer became confused0 This probably was due to losing fixation within the field and returning to the less efficient beam-canning behavior; at least in the writer0s experience9 $uch a tendency was compelling on the long durations of l-second flicker. Final consideration will be given to the effects of differing -searchlight separations from the observero In nearly all our experience9 increased separation from searchlight to observer results in improved target visibility up to about 75 yards separation. Beyond this displacement, there may be some loss In our experiments we did not have the effects of atmospheric light from the searchlighto For the conditions of no displacement between observeZo and searrChlight9 we h e obtained artificially high detection levels since under these conditions9 the observer must look through a substantial veil of back-scattered light. This veil would markedly reduce contrast for the target and also affect the observer visually (adaptation level9 pupil responses9 and "eyeballl" veiling glare) in a very adverse manner. While observing during Fort Knox field tests9 the writer was able to detect a tank at 500 yards by moonlight -lone, When looking at the same target along the searchlight beam (from the first adjacent platform) the target barely could be made out at allo The writer then attempted to assess the range of this interference by movitng to a platform about 10 yards displacedo From this van tage pointt the immlediately adverse effect of the searchlight seemed drastical ly reducedo Of course9 the extensive data of the field study bear more formally and d quately on these consideratifons. 31

The University of Michigan ~ Engineering Research Institute 2699- 1F From the observations described9 however9, it would seem that insofar as our illumination scaling was done correctly, the searchlight displacement findings are probably adequate except for the SL-C conditions since the task confronting the observer seems to be a complex identifieation task rather than a simple detection tasko In other words, most of the stimuli may be far enaogh above detection threshold levels under the conditions of our study that the back scatter from the searchl1ight would interfere seriously only in its immediate neighborhood. (Some further information of this point is treated in Appendix A0) As noted earlier9 because of limitations of time and relative emphlsls given to studying searchlight forms of illumination9 no study was glvin to flare illumination0 Gordon's research (Ref o 3) is directly relevant to flare conditions0 In general, he found that ~wher. the flare was directly behind either the observer o? target9 its visibilirt-. was best, As it was displaced to the $ide9 in az'misth9 visibility decrea'.~re a Comparisons cannot be made between -Gordon's and our study to assess relative performance under searchlight and flare illumination0 Such relationships probably are worthy of further investigation0 3~

The University of Michigan ~ Engineering Research Institute 2699-1-F REFERENCES 1 "Photometric properties of the atmosphere. Interim report on the brightness of, and the illumination from, the night sky" Great Britain, Ministry of Home Security, Civil Defense Res. Committee, Rept. RC(G)5, 12 p. 1942 (Confidential) 2 Pritchard, B, S., and Blackwell, H, R.'Preliminary studies of visibility on the highway in fog" University of Michigan Engineering Research Institute Report 2557-2-F, July 1957 3 Gordon, D. A. and Lee, G. B. "Model simulator studies of the night visibility of military targets as related to position of artificial illumination" University of Michigan, Engineering Research Institute, Rept. 2144-341-T (in press) 4 Blackwell, H. R. Psychophysical thresholds. Experimental studies of method of measurement University of Michigan, Engineering Research Bulletin No, 36, January, 1953. 5 Tanner, W. P., Jr. and Swets, JO Ao "A decision making theory of visual detection" Psychol Rev., 61, 401-409, 1954 6 Kincaid, W. M, "Theoretical models for the discriminatory process in visual detection" University of Michigaii, Engineering Research Institute Report 2144-281-T (in press) 7 Smith, So W,, Blackwell, H, R,, and Cutchshaw, C. M. "The effects of target size and shape on visual detection: III Effects of background luminance, duration, wavelength,, and retinal location" University of Michigan, Engineering Research Institute Report 2144-346-T (in press) 8 Gibson, E, J, "Improvement in perceptual judgments as a function of controlled practice or training" Psycholo Bullo, 50, 401-431, 1953,__ _ _ _.............. 33

TABLE I Results for M-48 Tank on Uniform Terrain, Observed Under Starlight Illumination C A, Response Frequencies. Correct Respoanses Wrog esponses Distance "Yes" "No" "Yes" Observer to Target Target Total Target, Target Total Observer Tarest (Yds) Present Absent Correct Absent Present Wrong Total N. C.S. 135 5(.50) 5(.50) 10(1.00) 0(0.0) o(o0o) 0 1010 180 4(..4) do) (0,9o) l(,10) 0(0.0) l(.l) 10.608 2 202 7(V35) 6(.30) 13(.65) 4(,20) 3(.15) 7(V35) 20.067 225 6(,30) 7(.35) 13( 5) 3(,15) 4( (20) 7(35) 20.067 B. Average Response Latencies (Seconds).. ft. CorrEct Responses Wrong Responses 2 Distance "Yes" "No" "'Yes" "No Observer to Target Target Total Target Target Total Observer Target (Yds) Present Absent Correct Absent Present Wroqg Total 14 f C.S. 135 io,4 18. -- -= 180 21.2 25.8 -- 48. 202 25.3 23.5 -- 20.0 33.3 -- -- -- 225 19.5 32.9 -- 27.7 28.3 -- -- - Report 2699-1-F

TABLE II Results for M-59 APC on Uniform Terrain, -I Observed Under Starlight Illumination A. Response Frequencies Correct Responses Wr s R Distance "'Yes" "No" —'.es" No -. Observer to Target Target Total Target Target Total Observer Target (d) Present Absent Corect Absent Present Wrong Total CoS. 135( 5(50) 5 (50) 10 0 0 0 10 o1 00 180 5(o50) 5(,5o) 10 0 0 0 10 1.00. 202 3( ~30) 3(~30) 6 2(.20) 2( 20) 4 10 ~ 037 225 9(.,30) llg(37) 20 4(o.l3) 6(.,20) 10 30.o90 270 4(.20) 7(,35) 11 3(.15) 6(e30) 9 20 o011 B o Average Response Latencies (Seconds)'. Correct Responses Wrong& Responses 1 Distance':Yes" "No" "Yes" "No" Observer to Target Target Total Target Target Total Observer Target ds) Present Absent Correct Absent Present Wro Total t~e ---- __ ~- I t S. 135 8.2 13.2 -- _o _ 180 9,8 30 2a -- 202 27,3 29.3 -- 295 315 -- = 225 2600 34 04 -- 4103 36.8 - 270 19 0 22 3 -- 26.3 1703 _o Report 2699-1-F

iIB~dE& ZEX:: r: TABLE III Results for M-48 Tank on Uniform Terrain., Observed Under Moonlight Illumination A0 Response Frequencies Correct Res ponses Wong eponses Dis tance "Yes" "No "Tes" "No",n Observer to Target Target Total Target Target Total Observer Taret ) Pesent Absent Correc2tAbsent Present rong Total M o CS o o 792 26( 50) 19(37) 45 6(o 13) 0 6 51 562 900 20( 50) 16(o40) 36 4(.10) 0 4 40 9608 oo1008 19(o46) 16(~39) 35 4(o10) 2( o05) 6 41 o409 1116 12(o40) 11(-37) 23 4(.13) 3(o10) 7 30 ~ 226 m 12214.8( o3) 7(~34) 15 4(o.8) 2(o09) 6 21 o143 14 Col a 900 l2(.40) 13(o143) 25 2(.07).3( 10) 5 30.346 1008 15(o38) 13(o32) 28 7('18) 5(912) 12 40 ol24 1116 9( 30) 10( 933) 19 5(.17) 6(o20).11 30 o050 WoD 762 4(,40) 4(.40) 8 1(o10) 1(910) 2 10.290 900 33( 37) 35( 39) 68 10( 11) 12(go13) 22 90 9210 loo8 19(~38) 21(~42) 40 4(,08) 6(o12) 0 50 283 1116 11(o28) 15( 38) 26 5(.12) 9(o22) 14 40.o74 Report 2699- -F

TABLE III (Cont'd) B, Average Response Latencites (Seconds) Coect ReLposs Wrong Responses Distance "Yes" No "Yes"'"No o Observer to Target Target Total Target Target Total Observer Ta! e.Yds_) Pre.sent Absent Correct Absent Present Wro_ Total M m e. 792 21 o7 497 39 o2 go00 36,0 60o1 54 o4 - 1008 30.5 55 - 59 5 45.6 1116 31.5 38.3 42 6 45,6 1224 26.9 318 33 9 34 C e.H g9 905 a26,1 13 5 39 3 m 1008 15 o4 347.7 7 22,8 1116 16,2 24 3 2- 24.5 24 6 24.6 W.Do 762 17e8 22.0o 24.8 25.7 - 900 20,1 34,5 = 35.8 37,9.... 1008 29.4 40 o.6 32 4 43.5.. 1116 35.4 43 4 -- 30.6 53 8 Report 2699-1-F

(1 TABLE IV Results for M.59 APM on Uniform Terrain3. 2 Observed Under Starlight Illumination A. Response Frequencies Correct Responses Wrong Responses Distance "Yes "No" "Ys" "No" f Observer to Target Target Total Target Target Total Observer Target KYds) Present At as ent Correct Absent Present Wrong Total C.1I 579 5(.50) 5(,50) 10 0 0 0 10 1400 792 4 (.40) 2(,20) 6 3(430) 1(0.lo) 4 10.001 0goo 7(35) 5(.25) 12 5(.25) 3(.15) 8 20.027 1008 5(.50) 4.(40) 9 1(.lo) 0 1 10.624. m 1116 2( o20) 3(30) 5 2(,20) 3(.30) 5 10.0!, W,*D 792 5(o50) 4(.40) 9 1(.1o) 0 1 10.608 900 15(,o5) 12(424.0) 27 3(.l0) 0 3 30 4608 1oo8. 12(1,40) 13(( 43) 25 2(,07) 3(.10) 30.346 u5 1116 6(2ao) 11(,37) 17 4( 13) 9('30) 13 30.021 CS. 792 5(.50) 5(o50) 10 0 0 0 10 1.00 1008 15(.so) 8(.27) 23 6(.2o) l(,03) 7 30.346 1116 9(.45) 8(.4.o) 17 1(o05) 2(,10) 3 20.412 12224 9(144.5) 45(.2) 14 24.(.2o) 2(.10) 6 20.127 Report 2699-1-F

TABLE IV (Cont'd) C B. Ave-rage Response Latencies (Seconds) Correct Responses Wrong Responses Distance'Yes" "No" "Yes" "No"s Observer to Target Target Total Target Target Totoa Observer Taget (YdS) Present Absent Correct, Absent Present Wrong otM C R. 579 3.03 15.8 792 120. 31.2 28.9 18.o -9 900 16.0 26.5 14.1 223.8 1008 9.6 29.2 33.6 1116 18.o 22. 3 11.8 30.4 W.D. 792 21.0 30.6 -- 18,6 - 900 25.3 47.0 49.1 1008 21.4 45.6 24.2 35.9 1116 40.5 50,1 40.7 39,7 CS, 792 26.8 31.2 - 1008 37.2 57.8 70.5 30.4 1116 28.4 59.0 -- 40,0 36,2 1224 27.6 30.0 36.0 42.3 42 Report 2699-1]-F

TABLE V ResAlts for Anti-Tank Gun and Crews Observed on Unifo-rm Terrain Under Moonlight Illumination A. Response Frequencies Correct Responses Wrong Responses Dis tance "Yes" "No" "Yes" "tNo",, Observer to Target Targget Total Target Target Total Obsert er Tar e( ) lPresenlt Absent Correct Absent Presenlt Wr Total WoDo 351 is5(o50) 15(50) 30 0 0 0 30 Lo 414 4(.40) 5(o50) 9 0.( 0o) I 10.608 CD 468 1a2( 40) 14( 47) 26 1( o03) 3( o10) 4 30.469 540 9(.30) 13(.43) 22 2(.07) 6(.20) 8 30 o7 m 612 10( 33) 9((30) 19 6(.2o) 5( 17) 11 30 o050 0 684 9(e45) 6(.30) 15 4(.20) 1(,05) 5 20 o220' CoSo 351 5(o50) 5(.50) o10 0 0 0 10o 1oOO 414 11(038) 8(,28) 19 6(o22) 4(.12) 10 29 057 468 a2(.40) 10(o33) 22 5(o17) 3(olO) 8 30.170 oh 351 20(40) 20(40) 40 o5(o0) (o) 5290 4114 1 3(.43) 13(.43) 26 2(.07j) 2(o07) 4 30 o 46 468 13(.43) 13(.43) 26 2(.7) 2g( 07) 4 30 o46;54o 9(030) 9(o30) 18 6(o20) 6(o20) 12 30 o037 - Report 2699-1-F

TABLE VI(Cont ud) ~CI B. Average Response Latencies (Seconds) Correct ResPonses Wrong Responses Dista nce es"1ye "No" "Yes" "No" o Observer to Target Target Total Target Target Tot*l 1 Observer Ta et (Ydu) Present Absent Correct Absent Present W g ToM. W.P. 351 14.4 21.2 - 414 20.0 36.0 34.9 w 468 27.1 40.9 77.4'. 37*9 2 543.8 35.1 46..1 612 12.3 30.2 21.0 29.4. 0 H684. 14.7 34.4 27.8 11.3 m C.S. 351 15.6 32.9 -- 4.14 21.7 29.4. 361.3 2427.6 2 o,~~~~~~~1 4.68 4.1.9 38.5 4.3.8 56.2 414 15.9 37.3 37.4 23.7 4.68 17.-3 31.3 25.4. 41.6 540 28.3 38,2 35,3 27,0 - Report 2699-1-F

TABLE VIC Frequency of Response Results for Anti-Tank Gun and eCre Obsrevd on Uniform Tetrain Under Sesxchlight Illumination (ObacPver.- COH.) Correct Responses Wron Response Ditance "YSit'No"'Yes"o Obarver to Taget U rget Tot*1 Tmget Twgt Tgt,*1 6%~ _k ~ T~tp~o_~~P~T Present Absent, Corret Ab s e-D % t Presen T~otal I SL 75yds R 8stc 1200 9: p45) 12050) 19 0 20(o5) I2 SL 75yd a R 48 1200 10( 50) 1O(e50) 20 0 0 0.20 LO r,)SL 75yda R 2Rse 1200 8W ko) i0(o( ") 18 0 2(4.10) 2 20.6L SL 75yda BR laea 1200 63~ 16 0 k4.20) 20 m Report~ 2699-1-F

TABLE VI I Results f or. M-48 Tank on Non-unifro m Terarin Observed Usnder Moonlight Illumination (Xidfield Position) A-. Response Frequencies Correct Responses wog sonses o Distance "Yes" o" "a No" Obser'ver to Tiiaget Targett Total T arget Taget Total Observer Target fYds) Present Absent Correct Absent PrBesenit Wrong Total COI{ I. 77!0(.50) io(,so) 20 0 0 0 20 ~ K00 U 594 12(.40o) 5(0so) 27 0 3( 0) 3 3.608 666 9(.30) 13(.433) 22 2(.07) 6(.20) 8 30 812 Correct Re-Spon'ses Wrong ResonsesDistance -Yes" "No" "Yesis" "No" Observer to Target Target Tot*l Target Target Total,Observer Taget Yds) Present" Absent Correct Absent Tresent Wrong Total N C.l. 477 5.6 18.o 591. 15.0 214.0 22.8 2 666 11.0 22.8 1-403 24.6 810 28.8 30.0 15,9 25 8 Report 2699-1-F

TABILE VII Frequency of Response Results for MH48 Tank on Non-unifof TeLr.rn Under Searchlight Illminati on (Upper Field Poition) s A0 Obsever CoHIo C Correct Responses Won Responses Di~ ~atce' *es.',Boe" "Noll" V'Ye" "INo": Simulated Observer to Target Target Total Tggt Taget Totlg IImntion T_ elt Yd) Preent Absent Corect Absent oPresent Wre_ Tot~ol SL-C ILere 1o08 5(o50) 5(050 ) 10 o: 0 1qo o SLS7yd$ R lec 5(050)0) o50) 0 10 Q0 0 0 10 0 S | Isge, F 2~se T 5( 50) 5(os0) 10 0 0 0 10 1o00 SL-C 2aec 1512 7(035) 10(o50) 17 0 3o) 3 20 SL-C 8e 9( 045) 9(o 45) 18 ( ~O5) ( o,) 2 20 0535 SL- 675yd8 1 2se) 6(,30) 8(1+40o) 14 2( c1) 4(p20) 6 20 I'27 rSL=75yds R 49ee 6(038) o( o50) 16 0 4(g o) 4 20 2396 SLo75yds R 8sec 7(035) 10(050) 17 0 3(o15) 3 20 492 m SL-i5yrds, Ro 15sec -9( o45) 9(o5)I 18 1(o05) i(o05) 2 20) o535 1see F 4sec T 10(o50) 10( 50) 20 0 0 0 20 lo0O o lsec F 2sec T 23( 38) 30(o50) 53 0 7(o12) 7 60 o562 8 1see F 15lee T 6(o30) o0(o50) 16 0 4(o20) 4 2a 4396.2 Bo Observer CoSo Correct RWong Re4 onse Dis tance "Yes" "No" "Yes "No' SimlaraXed Observer to Trget Total Target Totgel Tgotal llumina tion Ta rget (Yid) Present Absent Correct eAbsnt GeseGt Wro To SL-75yds R Isec 1512 6( o30) 0(6 4(o) 416 0 4( 20 o396 4 SLh75yds R 2sec 10(o5Q) 10(o50) 20 0 0 0 20 100 a Isec F 25ec T 15(~38) 18(253) 33 2(o05) 5(o.2) 7 40 o359 Report 2699-1-F

TABLE IX ftequency of Response Results fot M-48 Tank Obstrved onNoin-uniform Terrain Un-dr- Sarchlight Illumination (Ttee~t~e Position) Obsegvega C.H. c rgeet Wt g REes POn- e o B~~si~%~ge ~ egp~obsir vwt'to Utget T~gxr et To %*1 Urg t Ueget T ot*l ioo8 9 -) i OC.50) o9 (0 1 20 758 S $X-6 E ~24 1Y1.43), 1 (~ 26 20cf LT 2,3.416,3L-C 6rse', 12ag42 n4f 1350 7- 5) A 5 2k0 U 5)520 ~dL- 6 8~~6u 7~.35) 8(.4 ) 5 2K10) 2015) 5 20 SXL37yd- -C " 7(35 1.5J 7 0.15 20 m5 % 3 20~~~~~~~~~~~~~ SLW-37yds R c 5 )y.25j 5(.25) 1V0 5,2$,.25) 10 20.01 UI SL52 I 7(.35) 85(.20) 15 2(.20) r(.25) 5 20 1IeC F 1B~~bC 6(.30) )( 145> a ) 1SpS 1i.05~ 4(.20) 5 20.220 u2 SL75Yd8 R. 29e 1512 6(,30) -35 25) 3) 20.014 R SL-75Yds R 43ec 6(.30) 5(.25) 11 5( 25) 4(.20), 9 20 SL-75yds R 8sec 7(.35) 5(.25) 12 5(.25) 3(.15) 8 20.027 b SL-75yds H 15sec 44.20) 3( 15) 7 %(35) 6K -3r) 13 20,067 2R 2 H~~epo~ 2699-1NF

TABLE X F oequnc~y of Respons. Res$ults fo 59 M A Obseed on Non unifom Terain Under SeQ ch.light 1ll,~inb tion.Upper Field Posttioni A0 Obsegvew C or _ CoRrpect_ # _WR -ew o IHtstxe.@ it@8 flye. Sitm.lAteti, Obvtg@L tt Tfgbt Target Totl Tte aget Ttt Tot1 I l To.ntion T_) & sent Absent o A nt ftesenEt WK Tot1 o n SLT754sd R 2 se I512 7( 35) 9(~45) 16 1o05) 3(.15).20 290 SL e.5 d @ k 1512 6( o30) 9( ~45) 15 (0o05) 4(o20) 5 20 ~,2, 2 sL-o757 - 1 512 (A45) I 5O) 0 o(.o5) 1 o 5820 I ecg F 2 m e T 1l51 2 41 o;20) 0(. ei5) 1L 105;) 60 e3 0) 20 o " l.Xa< F -,e.-< F 1lW it s 7 ~35) 10(-50) 50 3( 05) 3 20 l 2(. F 4ec T 5 735) (o) 0 3(15) 20 492 m B. Obsever C.oSo.Corect Res. _ Wrog Re one )Dis tance "" "fNo R" Ye$" "No' Simalated Observer to Target Target Tota1 Tngeft Tiget Total l11.mintiton T t d) Present Absent Cortect Absent PrWent WrOg otq t SL-75yda R. c 1512 5(o33) 4(a27) 9- 3(20) 3(020) 6 15.034 SL-75yd$ R 2sec 1512 9(045) 6(~30) 15 i4(203 (g5) 5I 23Q p20 SLo750ds R 48ce 15a2a 21( 37) 14 40o) 23 3goa 1O 4( 13o ) 7 30 o226 3 Isee F 2ec T 1512 6(0 30) 7( 35) 13 3( o15) 4( o20) 7 20 067;' Report 2699 -1-F.~~~~.........

TABLE XI F-equen(e Y of %Sgeeportn RtaultR for M-a59 AK, O9bs@ wed1 onf No-~Uinf:nog Tgreean N ~ P5~I~~~nd S chl}e 3 hig ht Jlttu mirntion (Tr.see Positions) vis t *nogq OAYe8'a llNOF iswgerol ^ SimulatiAd Obsexvevg ta Tiget Trgget Tot~l Ttge~t T*egt Tot*lH | ~9SL-V 4$ist~ $1o08e 6F(8t3. o 84 f, oO 2C)) 6 a2'0 SL-0SC 4eo: fL T 1100 7.C5/ 8F 0" o1! o5 5i |L=75ydL s R k4< I~i0^0 10(5.;rp': 20 0 eO O -s |_ L3'. 4gecl21v 22: o2r? _ ) 18 4o ~o0 =e s ~-"'.? 5 Ul O5 t 6. o175, ~ t;> 26 r 7.g x; g v Q F',I I WL j'J 5 4 Na5.e,~:S aa 9[ ~3 ] 18 ooL 26 C SL75 R 475 S78 L( o7 II51 7'2,,,.,2I ~... SL 150 R i 175.18( 4oi45..3,): 7 o3 0c 26I 2 70i 3o 40 3 SL~7c 4.e Li, o'y,0): o387f L, ('', 5(.5: oo20 ) 60 o-a S~ Ib-C o065^R.1e175 L 8,KP 11 72 45 6 0 20 3( 2) 9 20 SL k: 8e 11.7, 5(o25) 6(.3o,) 11 4(oaO) 5c 25) 9 20.1 m SL 75 R 8e9c 1175 9(o4 5)' ~0( o5) ~ 9 o oo) 0 a1 ~.0 1 20 _. l~,::c. F 4..e < T 1175 1.8( 05) 516)o40 34 4(Io) 2(.05) 6 4i o.2 = ei:cz, F.8L6. T 17 c5 8L o 4) 7( 35) 15 3( 15) a2(o ) 5o 20 1 2a.= F 0k e. 0 2'1~r 28()o4 5 o 4,) 038 2 o5) 2 40t o7558' 2at i F 8 stc T 1175 10(50 ) io ( 50) 20 0 0 0 20 1t,0 0 SLoC 4,e e242 4(o20) 3(o 5) 7 7( 35} 6og30) 13 20.t. SL 3'2 R k~eo 2. a10(o25) 9(o23) 19 (o27) 1o0( o251 ) 2! 4: SL 75 / 27):IP g33 24 7I2) 93 2e'tRi 16 4o oO 1 SL-o 8&ee- v124 O2 4( 20)) 2( 0lO 6 8( o40) 6( 030.) 14 20 2 c S 5 A 4ae,- i350 6( o20) 0(6 o3) 16 5( o7) 9( o30) 1 30 o573 r SL.I0 R 4.13 l50 \ oIO) 4( 4oko 5 I.Ioo) 4( Jo) 5 Io o0o - S L` 5 R. 8& * 1350 5(.25) 8 40 ) 13 fcQ0u).,2 5 20 0 SL 15i. 81 e350 a{2 o20) 3230) 2.) o202) 3 0 20 2sRc- F 4sae T 1350 0( 25 2 0 22:2, 1 0(25 18 4 "0 = sEF e 1 350'7.... T o5 7 14 3 15) 3L o05) 6 20!i7 6 1epo~rt 2699 Cb1F

rI TABLE XII Frequency of Response Results for Antio-Tank Gun and Crews, Observed on Nonaeni-form Terrain Under Searchlight Illumination (Upper Field Position) Observer C H. Correct Responses Wrong Responses n Distance'Yes$" "No " "No" SimWlated Observer to Target Target Total Target Target Total Illumination Target (Yds) Present Absent Correct Absent Present Wrong Total Mo t-SL 75 R lsec 910 8(.40) 6( 30) 14 4(.20) 2(.10) 6 20 o27 COSL 75 R 2sec 910 9(.45) 10(o50) 19 0 l(.05) 1 20o 758 SL-C 4sec 910 3(,15) 9(.45) 12 1(0oO5) 7(V35) 8 20 007 m SL 75yds R 4sec 910 9( 045) 10 050) 19 0 ( ) 1 20 758 1 sec F 4sec T 910 10(o50) 10.50) 20 0 0 0 20 loOO' 1 sec F 2sec T 910 7(.35) 9(,45) 16 1(o05) 3(-15) 4 20 o323 SL 75yda R 8sec 1116 8(.27) 11(. 37) 19 4( o13) 7.23) 11 30.054 Report 2699-1-F

The University of Michigan ~ Engineering Research Institute 2699-1-F APPENDIX A In order to relate the model simulator studies better with field test data, documentation was obtained for the inherent luminances of targets and their backgrounds as used in the field tests. This documentation, carried out by M*. Pritchard of these Laboratories, was obtained using the photoelectric telephotometer to measure directly the luminances of targets and backgrounds under the field conditions employed. Two:points of comparison may be made from this documentation. Firstt, the degree of correspondence for levels of illumination established for the model may be compared with that occurring in the field. Secondly, comparison may be made of relative contrasts.for selected portions of the targets and backgrounds. With regard to the first comparison, above, only two sets of data from the field tests provide the desired information. This was because atthe time documentation was obtained in this regard, following the actual tests, a limited number of target conditions had been studied when the last remaining 2500 watt searchlight lamp burned out. The remaining documentary data'were obtained using a 2000 watt lamp and although, the contrast information is of use in comparison with model simulator condftions, no absolute comparisons are possible. The sets of documentary data obtained when the 2500 watt lamp was available provide the comparisons of inherent luminances. appearing below. A. Tank, 1000 yards observer Post 6 (separated from searchlight approximately 140 yardsj (luminance in foot lamberts) Model Study Area Measured Field Study.(SL75 R) L Track.0090.0154 Center.0047.0088 R Track.0104.0183 Turret.0058.0150 L Bkgnd.0106.0170 Tank Shadow.0029 o0015 R Bkgnd.0o104.0148 Top Bkgnd.0022.0150 Bottom Bkgnd.0018.0154 49

The University of Michigan ~ Engineering Research Institute 2699-1-F B. Tank; 1000 yards; observer Post 1 (adjacent to searchlight) Model Study Area Measured Field Study (SL-C) L Track.0780 o0162 Center.0685.0134 R Track.0744.0244 Turret.0658 o0l146 L Bkgnd ~0780.0185 R Bkgnd o0780.0150 Top Bkgnd.0560.0150 Bottom Bkgnd.0780 o0166 Examination of the foregoing information shows, first, that fairly close correspondence results when measures were taken with the searchlight displaced at its maximum distance from the observer post. The luminance values shown for the model target are approximately twice corresponding measures taken in the field. It is possible that greater accuracy could be achieved in the model study in arranging the searchlight beam to center exactly on the target and illuminate it with peak candle power. This possibility, at least, is consistent with the difference noted. When the measures were taken with the searchlight very little displaced from the observer post, shown in the second set of data above, the luminance values obtained for-the field study are about a half log unit greater in each instance compared with corresponding measures for the model simulator. This result is indicative of the effect of light backscatter at positions adjacent to a searchlight. The telephotometer, in the field, received a considerable amount of back-scattered light which increased the apparent luminance for selected target areas, and, as well, reduced contrasts between these areas. Due to the absence of scaled atmosphere in the model studies, these effects are not seen. Although the luminance values for target areas are higher, they are very little higher for the SL-C condition compared with the SL 75 R condition. The field data obtained using the 2000 watt lamp may be used to evaluate the range of effect of searchlight back-scatter on contrasts. For this purpose; the luminance values obtained at 500 yards for the front center of the APC and also the sloping portion of the front are listed below as measured at' the first five observer posts. Post 1 is adjacent to the searchlight.o Approximate separations for the other locations are shown in parentheses. Obs, Post: o lOyds ) i ((20yd s) #4(40yds) #5(80yds) Slope Front.100.023 o017 o013.011 Center Front.097.023.022 o.018.013 Ratio ("contrast. 1.03 1o00 1.12 1.14 1.12

The University of Michigan ~ Engineering Research Institute 2699 -I-F From the preceding analyvss, it may be seen that the high increase in luminance for target areas due to searchlight back-scatter is essentially gone when one moves from Post 1, adjacent to the target, to Post 2, ten yards removed. The remaining values reflect some further change, but of a much smaller magnitude o The"contrastse noted between the two selected target areas increase with greater separation of observer and searchlight, and cease to show back-scatter effects at Post 3, 20 yards from the searchlight0 In the model studies, the searchlight displacement labeled SL-C would not have the realism of field conditions due to the absence of scaled atmosphere and resulting absence of back-scatter. However, if displacement of 10 or 20 yards is sufficient to obvi:ate back'scatter effects, our other conditions should otherwise constitute adequate simulation0 To get a better idea of the correspondence for target and terrain contrasts between the model and field conditions, the following comparison may be made0 The set of field data for the tank measured at observer Post 5 (80 yards displaced from searchlight) may be converted to relative values by determianing the ratio of each area measured to the center area as a reference point. This leads to the values to be shown shortly0 Similar treatment is given to measures on the model simulator for the tank at the equivalent distance for the similar displacement of the searchlight, These values are shown below for comparison with the first set. Area Measured Field Tank Model Tank L Track 1o55 1.66 Center 1. o00 1 o.00 R Track 1o55 2.00 Turret 1o12 1.66 L Bkgnd 2.00 1.89 Tank Shadow.78.02 R Bkgnd 2o00 1.66 Top Bkgnd.89 lo24 Bottom Bkgnd 1.55 1.66 It would appear that the internal contrasts for our tank target are a little greater than for the tank in the field. Also, contrast for the tank with the background at the top and bottom terrain areas is greater for the model. Terrain at the sides does not form as high a contrast. In the field, terrain ahead of the tank sloped away from the searchlight and the woods behind the tank, although in its projected background, were quite distant. Our terrain was simply more uniform in a relative sense.. The contrast of the tank shadow and adjacent parts of the terrain and target is much greater in the model situation than in the field. However, in the field the 2000 watt searchlight lamp provided less illumination on the target and, at the time of measurement, a full moon was illuminating the terrain alsoo The luminance of the shadow, in the field measurements was o.002 foot lamberts which could be due to moonlight primarily and possible some space light present. 51

The University of Michigan ~ Engineering Research Institute 2699-1-F The field study data provide a basis for much more extensive comparisons than can be attempted here0 With the exception noted, the correspondence between model and field conditions appears, in physical terms, reasonably close0 In the text there are several points of comparison in terms of of observer performance as well0 An examination of all of these poinlts of comparison is necessary to deterine how much confidence can be given to the extension of the model simulator findings to field conditions.. _. _ ~~~~~~~2

The University of Michigan ~ Engineering Research Institute 2699 — I F APPENDIX B The deivation of t he quation M s hown in the text is quitestraight forwardo We begin 79ith the st'atament of several ter$m And relationships as follow v (1) I(S) = initial degree of un;ce'tainty in S (2) I(R) = degree of uncertainty in R (3) SI(R~) d $egre of unc et. inty in cobi ned R and. S (4) IR(S) = degree of nmcertainty in S given R W1,tht (1) and (g4) we he ve (5) I(S) R(gS3 extent of rd tion o I(S) given knowledge of R (equivalent to t:mulus information ained by knowig. R) Thens from stand rd workS on information theory: (6) I~(R)' - B s) + ia~) Adding I(gS) to each ide of e quation (6) and chPanging signs leads too (7) I(S). IR(R) + I ~s) (R9s) Note that the left hand side of (7) is the defined term (5)~ Also, because in our experiments,9 on a given observation trial the stimulus was equally likely to be present kor absents we know. that I(S). = I bit. Hence a solution for the right had'side of (7) evaluates I(S) - IR(S) and gives us the stimulus information gained by knowing R0 To accomplish this,, we substitute for I(R) and I(R$S) as folilow using response data to obtain the paticula pt probabilitieeso (8) I(R) = - ) log2 P(Y) + P(N) log2 (IN)] and gs ) log2 P(S Y) + P(S N) log2 I (9) I(RS) ( i.s') +og Ps, )g +(,S), soga Ns, ) + ls2 Y) 0og2 substituting from (8) and (9) in (7),'ewriting I(S) as 1, and using the appropriate constant to convey t from logarithms base 10 to base' 2, we haveo (o) 0 s) T(S- ) IS 1 A 32 V-(Y) log F(Y) -( P() log P(N) + PgSY) log F((S Y) + P (S 9N) log PF(S N)3+ P~ s Y~) log F(S' YS ) + P(s$' N) log @S' N)] Desig8nating I(S) IR(gS) with the symbol Mp a the measure of stimulus informatio n gained by knowaing R whae the formrnula cited in the text. 55

The University of Michigan ~ Engineering Research Institute 2699 I1F;'t, e, aftures of this relationship are easily shown0 If an observer always responded "yes"i, we would of cours gain no stimulus informationo In terms of the above equation, PF(Y) and P(N) = 0 which gives us, M = 1 + 3032 o-0-0 + 1/2 log 1/2 + o + 1/2 log 1/2 + 0] = 0 Also, if the observer is right (SY~ nd (S N) as often as wrong (S, Y) and (SN), M = 0. If he makes no wrong responses, maximum stimulus information is gained asrd M = 1. (This would be true if no right responses were made, also the system needs some external knowledge of accuracy in order to decode the responses ) Since the measure M was derived to provide a basis for avoiding the ambiguous effects of criterion differences in the yesano procedure used, it would be well to illustrate the relative freedom of this measure with an example. Consider two sets of data0 The target has been presented 10 times during 20 observation trials0 In the first, an observer has given the following responses"Yes" Target "No"' Target "Yes" Target "Nd' Target Present Present Absent Absent Frequency 10 0 5 5 A second observer gives these responseso "Yes" Target "No" Target B"Yes" Target "No" Target Present Present Absent Absent _-.._:.., _,,,.:.,-._,..:.~ _...- Frequency 5 10 0 5 It may be noted that the first observer has responded "yes" 15 times even though the target was presented on 10 times during the 20 trials. The second observer, on the other hand, has been willing to respond "yes" far less frequently0 The first has a high false positive rate- the second a low rate~ The proportions in each of these examples are treated first, by the conventional method0 The detection proportions (from "yes", target present responses) is adjusted in terms of the proportions of false positive respornes ("yes', target absent)0 Usually this is done by employing the following relationship: i-C Where~ PO = corrected detection proportion P = raw detection proportion C = proportion of false positives 5h -

The University of Michigan ~ Engineering Research Institute 2699-1-F In terms of the above equation, the first data set would have to be repre -.. sent-ed by a orreted detection proportion of OO00o The second would yield the corrected detec tion proportion of o50o If we analyse the same two sets of data for the stimulus information gquantity M. we would find that not different but the same values would result. For each data set M would be 0o31. Thus if, as is possiblen the sets are from two different observers in the same situation, one is not providing more stimulus information than the other0 It might be the case that the first observer is highly willing to give false positive responses whereas the second is highly unwilling to do so0 The freedom of the data measure. developed in information terms, from effects of such criterion differences becomes fairly obvious0 It was not possible to analyse the information statistic in terms of sampling and bias0 Since in a series of observations the target was presented on 50% of the trials, these question are probably not serious. In any event9 our measure is precisely defined aad appears to provide for less ambiguity in interpreting results 55_

~...i, ~:::~ aI ell. V1~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1~ ""~.........................'. Figure 1. The model simulator.

Off.............f; 0::::: - --: I' t:::S f::il:r 0; id: a:: V V1hi:-::~~i:: F igu re lll iSl iioi! i iii ill i -~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ i i~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -q- Figure 2. Experimental room and equipment.

oo i - _i............. ~!i i~" ~"""~.~.............'~i~'~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~...... 0n E IC _ _F" _?I; e........ 3..M-48 tank scalemode1. F igu re 3. M- 48 tank scale model1.

:i I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~ii~~?i ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~10...........................?ii.1~........~~~~~~~~~~~~~~~~~~~~~~~~~~~.............. F igure 4. M-59 APC scale model1.

ii$s 0\ r::::::i: O ~:::::~ T:,lj,:::::: I:: a;:I:::i::-: s -c~ 6 " :;:-:: r::::b ~::: rr:"::::: -: I~~I &i *:-::: —:I:::::: %i-:~: iM I~:ii:: *;_::::_.::::, i: *liia,liIsD:~~,ri ::i:::::ig:i:I,.II;ii:i 9 c Bi-s i:::::r:; -i, is,r:,;~"rp::g::;I:l-::::::i'i~::::::::i i:':':i::::;-:::i:i:I::::::::;li':i:::: Figure 5. Anti-tank gun and crew scale model.

0 r. TARGET ~ TALL w FIELD ~ ( GRASSl OBSERVER Figure 6. Map of original target positions during exploratory experiments. 61

GA-2002 HILLS. * "MIDFIELD" TARGET POSITION TARLIGHTS TARGET POSITION v OBSERVER Figure 7. Map of final uniform target field and target positions. 62

I -_ AHILLS ~I~REE' POSITION /(ZUPPER FIELD' POSITION 0 O:~ ~.IDFIELDO POSITION LOW OBSERVER Figure 8. Map of non-uniform target field and target positions. 63

1.0 1.0,M) OBS: C.S. w cu.8.8 z w 0 I I U) I CI W.6.6 < | /YES" TARGET PRESEIT Z 4 *NO" TARGET 40 0 0 tIABSENT I ~_ I —u Io 01 0 0.2.2 LL QM Z "NO" TARGET PRESENT AYES" TA ET 0 AB ENT 0 130 150 170 190 210 230 250 TARGET DISTANCE (YDS.) Figure 9. Results for M-48 tank, observed on uniform terrain under starlight illumination, showing relation between frequency of response data and information measure M.

1.0 *mP~ M-48 TANK.8 @-4 M-59 APC OBS: C. S..6 M.4.2 0 130 150 170 190 210 230 250 270 TARGET DISTANCE (YDS.) Figure 10. Results for M-48 tank and M-59 APC, observed on uniform terrain under starlight illumination.

1.0 OBS-0. C.H. C.S. - - W.D..8.6 M.2 740 820 900 980 1060 1140 1220 TARGET DISTANCE (YDS.) Figure 11. Results for M-48 tank, observed on uniform terrain under moonlight illumination.

OBS' 1.0- — * C.H., — C. S. 3 —-.I W.D..8.6 M.4.2 580 660 740 820 900 980 1060 1140 1220 TARGET DISTANCE (YDS.) Figure 12. Results for M-59 APC, observed on uniform terrain under moonlight illumination.

I.0 OBS:.- --- C.H. A- C.S..8 mm W.D..6 M 360 400 440 480 520 560 600 640 680 TARGET DISTANCE (YDS.) Figure 13. Results for anti-tank gun and crew, observed on uniform terrain under moonlight illumination.

1.0.8 OBS: C.H..6 M A 480 520 560 600 640 680 720 760 800 840 TARGET DISTANCE (YDS.) Figure 14. Results for M-48 tank, observed on non-unif orm terrain in midfield position under moonlight illumination.

1.0.8 (SL.C.r \ (4-SEC. OBS: C.H M.4 0O~0 0 1000 1200 1400 TARGET DISTANCE (YDS.) Figure 15. Results for M-48 tank, observed on non-uniform terrain in tree position under 4-sec searchlight illumination along observer's line of sight.

1.0.8 Q$St C.H.6 1350 YDS. M SEC.C DURATION.4.0 I SEC. F.2 4 SEC. TOTAL TIMEre 8 SEC. DURATION C 20 40 60 80 100 120 140 160 SEARCHLIGHT DISPLACEMENT (YARDS RIGHT OF OBS.) Figure 16. Results for M-48 tank, observed on non-uniform terrain in tree position illuminated by searchlight at differing displacements from observer.

1.0 OBS: C.H. / N SL 75 R / N\ 1512 YDS.. / N.8 / N / (I SEC. FLICKER).6 N N TARGET IN UPPER FIELD POSITIONN M | PO,.4.2 TARGET IN TREE POSITION 2 4 8 15 DURATION OF SEARCHLIGHT I LLUMINATION (SECONDS) Figure 17. Results for M-48 tank, observed in different non-uniform terrain positions under diffe ring durations of searchlight illumination.

SL 75R I.0 2 C 1SEC F XI\ OBS' C.H. 4 SEC TOTAL.6 A I SEC F A~r~o~c 4 SEC A~~~~~~~~~f TOTAL.2 1-SL-C 0 1000 1040 1080 1120 1160 1200 1240 1280 1520 1360 TARGET DISTANCE (YDS.) Figure 18. Results for M-59 APC, observed on non-uniform terrain in tree position under h-sec searchlight illumination at differing displacements from observer.

OB S 1.0 OBS: --- C.H. i —A C.S..8 SL 75R 1512 YDS..6 M I|pl SEC. F.4 / t // A/ SEC. F 1 2 4 8 15 DURATION OF SEARCHLIGHT ILLUMINATION (SECONDS) Figure 19. Results for M-59 APC, observed on non-uniform terrain in upper field position under differing durations of searchlight illumination.

1.0 OBS: C.H..8 - 4 SEC. X-K-~ 8 SEC. (1175 YDS.) M.2. 0 t C 20 40 60 80 100 120 140 160 SEARCHLIGHT DISPLACEMENT (YARDS RIGHT OF OBS.) Figure 20. Results for M-59 APC, observed on non-uniform terrain in tree position illuminated by searchlight at differing displacements from observer.

Figure 21. M-48 tank on uniform terrain (mid-field position) under moonlight illumination.

Figure 22. M-59 APC on uniform terrain (mid-field position) under moonlight illumination.

Figure 23. M-59 APC on uniform terrain (mid-field position) under illumination by searchlight 75 yards right of observer.

\O Figure 24. Anti-tank gun and crew on uniform terrain (mid-field position) under illumination by searchlight 75 yards right of observer.

Figure 25. M-48 tank on-non-uniform terrain (mid-field position) under moonlight illumination.

Figure 26. M-48 tank on non-uniform terrain (upper field position) illuminated by searchlight 75 yards right of observer.

Figure 27. M-48 tank on non-uniform terrain (upper field position) Figure 27. M-48 tank on non-uniform terrain (upper field position) illuminated by searchlight 150 yards right of observer.

co Figure 28. M-59 AFC on non-uniform terrain (upper field position) illuminated by searchlight 75 yards right of observer.

Figure 29. M-59 APC on non-uniform terrain (upper field position) illuminated by searchlight 150 yards right of observer.

Q o0 600 YDS. Figure 30. Target and terrain luminances for the three scale model targets when illuminated by searchlight along observer's line of sight.

.,?".rto' 0': ~~~~~19 ~ ~ ~ ~II cD o ONI 1000 YDS. Figure 31. Target and terrain luminances for the scale model vehicle targets when illuminated by a searchlight 75 yards right of observer.

100 400 CANDLES = —;HORIZONTAL 80 x —— x VERTICAL a: 60 w IJJ a.. IJ.. 0 z 40 cr a. 20 10 0,,//1 I I,.. I I I I, I 0 2 3 4 5 6 7 8 9 10 DEGREES Figure 32. Candlepower distribution for simulated searchlight beam using corrector slide.