ENGINEERING RESEARCH INSTITUTE THE UNIVERSITY OF MICHIGAN ANN ARBOR Final Report THE EFFECTS OF CERTAIN PSYCHOLOGICAL VARIABLES UPON TARGET DETECTABILITY H;. Richard Blackwell Vision Research Laboratories ERI Project 2455 BUREAU OF SHIPS, DEPARTMENT OF THE NAVY CONTRACT NO. Nobs-72038 WASHINGTON, D. C. June 1958

w~, ~.l (,k k P.25 (q -~ —

The University of Michigan ~ Engineering Research Institute 2455-12-F TABLE OF CONTENTS Page List of Figures iii List of Tables iv Summary v Io General Introduction 1 IIg Experiment I: The Effect of Knowledge of Time of Occurrence and of Target Size and Duration Upon Target Detectabi lity III, Experiment II: The Effect of Knowledge of Target 12 Location Upon Target Detectability IVo Experiment III: The Effect of Observer Criterion 17 Upon Target Detectability Vo Experiment IV: The Effect of Frequency of Target 21 Occurrence Upon Target Detectability References 26

The University of Michigan * Engineering Research Institute 2455-12-F LIST OF FIGURES Number Tit le 1' Contrast Factors for Absence of Knowledge of Target Location 2 Contrast Factors for Different Observer Criteria -__ _ _ _ _ _ _ _ _ iii

The University of Michigan ~ Engineering Research Institute 2455-12-F LIST OF TABLES Number Title Page I- Detection Probabilities for Frequent and 27 Infrequent Target Presentations

The University of Michigan ~ Engineering Research Institute 2455-12-F SU$MARY Four studies have been made of the: effects of psychological variables upon visual detection thresholds for simple circular targetso These studies were intended to evaluate the significance of certain differences which obviously exist between the conditions under which visual detection is studied in the laboratory, and the conditions existing in military visibility problems. The effect of these differences is in each case expressed in terms of the contrast factor, that is the factor by which target contrast must be multiplied in order to compensate for the presence of the difference. It was found that when a target appears without previous warning, a contrast factor of 1.40 is required to compensate for the resulting loss in visibility. Absence of warning that a target is to appear, and absence of prior information concerning the target's size and duration of appearance require a contrast factor of 1.49. The absence of knowledge concerning the precise location to be occupied by the target requires a contrast factor of at least 1.31 even though the observers know precisely when to expect the target. This variable requires further study, since the target in these experiments could always be expected in one of two possible locations. Analyses of a number of previously reported studies suggest that what is called a "commonsense criterion" of seeing results in a contrast factor of 2.40 in comparison with the laboratory method of temporal forced-choice. Finally, it was shown that reducing the frequency of target occurrence from one presentation on the average every 30 seconds to one presentation on the average every 15 minutes could be compensated for by a contrast factor of 1o19. Further studies should be conducted before extensive practical use is made of these factors, but they should serve to indicate the magnitude of the effects upon target detectability produced by psychological differences between laboratory and field conditions such as these.

The University of Michigan ~ Engineering Research Institute 2455-12-F I o GENERAL INTRODUCTION A number of experiments have been conducted in these laboratories since 1946, designed to provide quantitative data on the visual detection capability of the eye for use in predicting the visibility of military targets. These experiments have been conducted with experimental procedures especially developed in order to maximize reliability and efficierncy of experimentation in the laboratory. These procedures include rigid control of the physical characteristics of targets, and the use of psychophysical methods which have been shown to be comparatively free from extraneous psychological variables which might otherwise produce variability, and uncertaint-yin the use of the data. These methods place the laboratory observer under some unusual restraints. and require that he indicate his ability to detect the presence of targets under study within a rather unnatural frameworko It is of considerable interest to ascertain to what extent the data obtained in the laboratory under these conditions can be directly applied to practical visibility problems in the field, or to what extent conversi:on factors are required to compensate for differences which exist between the conditions enacountered in the laboratory and those prevailing in the field. The present report summarizes four experiments designed to evaluate the effect of differences which obviously exist between the laboratory and usual field conditions. Two of the experiments are concerned with possible effects due to the fact that in the field the observer seldom has complete information in advance as to the time or place of occurrence or the size or duration of the target, whereas this information is normally provided in laboratory tests. One experiment is concerned with possible effects due to the fact that targets seldom occur frequently in the field, whereas in the laboratory targets are usually presented in rapid succession, in order to make'efficient use of the observers. The fourth experiment is concerned with the effects of differences in the criteria used to signify detection in the laboratory and in the field. For some years now, we have been describing the effects of differences such as are here considered, in terms of contrast factors, That is, we express the effect of a given variable in terms of the increase or decrease in target contrast required to compensate for its effects$ This policy not only makes it easy to evaluate the relative importance of various effects. but permits a simple allowance for these effects to be made in the computational techniques used to predict visibility distances for practical military targets. Thus, the intent of the present report is to evaluate the contrast factors rbquie to compeensate f aor arious pychological vfariables which obviously difafer between the laboratory and the field.

The University of Michigan ~ Engineering Research Institute 2455-12-F Since the four -experiments were conducted with different apparatus and procedures, it has seemed appropriate to describe these experiments in separate'sections of the present report.

The University of Michigan ~ Engineering Research Institute 2455-12-F I I BPEL*RIEN I The Effect of Knowledge of Time of Occurrence and of Target Size and Duration Upon onTarget Detectability* A Introduction This was the first of two experiments in which effects due to restricting the information available to the observer about the target were studied. In the usual laboratory situation, the subject is given nearly complete information concerning wheroe and when to expect the target, and the size shape and duration of; the target presentation. Since almost none of this information is normally available to the military observer, it is importantto assess the effect of the amount of information available to the observer in order to compensate for any effects found in applying laboratory data to the solution of practical visibility problems. Experiments were conducted to investigate the general order of effect due to restrictions in various kinds of target information. Since the experiments were exploratory., they were conducted with quite crude apparatus and procedures.o The data were collected, however, under conditions intended to facilitate adequate comparisons between conditions differing with respect to the information. variable0 B. paratus and Procedures The apparatus used for these studies consisted primarily 6f a translucent plastic screen which was illumnated diffusely from in front to provi3de a background of uniform high luminance, and which was illuminated from behind by a projector to produce a comparatively small target. The diffuse illumination was provided by constructing a white cube in front of the screen, with a large opening in the side opposite the screen through which the observers viewed the screen. Frosted lamps were mounted in the cube in positions which were invisible to the observers. The luminance of the screen was maintained at values ranging from 40. to 160o foot-lamberts in different experiments. The projection apparatus was built around, a Bausch and Lomb Balopticon projector. A wheel was inserted in a plane near the condensor lenses so that metal apertures mounted in the wheel could be imaged by the projection lens on the rear of the translucent screen with uniform l uminance. A timer wheel was placed near the plane of the projection lens so that the targets would come on and fade out uniformly as the timer was used to present the targetso A flag shutter was inserted in the beam near the plane of the projection lens which could be manually operated to permit presentation of a target whenever the opening in the timer heel came past the projection len$s Thus, the temporal 8character istics of the target presentation were controlled by the characteristics *This experimelnt was conducted by Russell L De Valot$s, 0o Thomas Law, and Wilson P. Tanner, JrQo

The University of Michigan ~ Engineering Research Institute 2455-12-F of apertures in the time wheel, rather than by the mode of operation of the hand shutter0 Targets of varying sizes could be presented on successive trials, since apertures of different sizes were mounted in the target wheel, There was room for eight different apertures, Targets could be presented for different durations, since several different apertures could be mounted around the rim of the timer wheel. There was room for only three different timing apertures. The wheel bearing the timing apertures could be rotated with respect to the drive shaft of the timer into one of four positions. This rotation was accomplished by holding the wheel while the shaft turned with respect to it, Once rotated, the wheel bearing the apertures was carried along by the drive shaft firmly and continuously. By manipulating the position of the wheel with respect to the shaft, the experimenter could select which of the three exposure durations would be in place at the aoment he wished to withdraw the flag shutter and expose the target. In addition to all these wheels, there was a wheel containing neutral density filters of varying transmittance which were used to vary the target contrast within the threshold range, The experimenter rotated this wheel to the desired filter between successive exposures of the target. This multi-wheeled apparatus was manually operated by a team of two or three experimenters in accordance with the requirements of a pre-established presentation schedule. There was no way to record the $ettings of the apparatus actually made but it is believed that the few errors which tmust have been made had only a trivial effect upon the data from the experiments. The observers utilized the'yes" - "no" method throughout these experiments, recording their "yes" responses only. A "yes" response was signaled by the observer by pressing a doorbell buttonr The button activated a circuit which in turn activated one pen of an EsterlineAngus continuous paper recorder, Another pen was used to record periodic time markers. Since the experimenters worked with a rigid time schedule, it was possiable to evaluate whether the recorded positive responses on the paper record did or did not correspond to the presentation of a target. The paper records were manually scored and the correct responses and "false alarms" were tallied for each class of target being presented during a given experimental session. There were basically two experimentsb In the first, five conditions were studied representing different amounts of restriction:in the information made available to the observMers Description of these experiments can best be given in conen@ection with the presentation of the results in Section III below, In the second experiment, one grand comparison was made beltween two extremely different amounts of information restriction, Here agaitn, the procedure may bes$t be described in connection with presentation of the results.

The University of Michigan ~ Engineering Research Institute 2455l12-F C, Results Experiment I The background luminance was maintained at 160 foot-lamberts for this entire experiment. Condition. 1. In the first condition, target presentations were made follow'ing by 2 seconds ab warning buzzer which served to alert the observers and to define nearly exactly when the target was to be presented. A circular target subtending 99 minutes was presented for a fixed known duration of 1,5 seconds once every 28 seconds, The probabilities of "'yes" were corrected for "guesses" in accordance with the relation: p' = P-'. (1) 1.V where pl = corrected probability; p = raw probability; and V = probability of false alarms The observers were practiced under these conditions for eleven experimental sessions of two hours each before they were considered trained, Two experimental sessions were conducted before the other four experimental conditions were studied, and two more sessions were conducted after pll other conditions had been completed, Threshold contrast values were -estimated by visual fits of ogive curves to the experimental probabilities, Contrast threshol1dsare presented below for three observevrs, representing averages of the values obtained in the four experimental sessions at the beginning and end of the experiment. Observer Threshold Contrast S 5 0171 Ns o0120 B.0091 Average.0127' Condition 2 Nine session were conducted in which the observers were given no warning as to when a targetL was to be presented. The intervals between presentations Were of varying lengths, with an average interval of 28 Seconds sO that.the frequency of target presentations was equivalent to that used inl condition 1, In each ses$$ion only one target size and one target durlation were used and the observers were shown the target in advance at suprathreshold contrast, so that they knew exactly what kind of target was to appear~ Three sizes of target were studied as follows: 73, 99, and 121 minutes, Three durations were studied as

The University of Michigan ~ Engineering Research Institute 2455-12-F follows: 1, 1 5o, and 1.9 secondso One experimental sessiona- was devoted to each size.for each exposure duration0 (It should be noted that both the size and duration ranges used were not large, due to the requirement that the thresholds were to be essentially equivalent for subsequent e''erimental conditions to be possible. However, the phenomenal appea.ances of the nine targets used were considered markedly different by the observers$o) In analyzing the data from this conditionL,, and the other conditions in which no warning was given the observers, the raw probabilities of "'es" responses were corrected for false alarms by means of- equation (1). However, we must consider the method used to define V. In the usual experiments in which the observer is warned that a target is about to appear, he is expected to res@pond within a few seconds and false alarms are clearly defined by.'those responses coming immediately after the warning when no target was presented.l When there is no warnlng, the observers may respond at any time, and it is somewhat ambiguous what is considered a false alarm and how a probability is computed~ We decided to divide the entire experimental session into 2 second time intervals'. Any "yes" response within 2 seconds of a target presentation was co idered correct; any other response was considered a false alarm, Each 2 second interval was considered to be an opportunity for a false alarm and the percentage of false alarms was computed accordinglyo Actually there were very few false alarms so that the corrections for V were seldom of any significance0 At this point, we will consider only the data obtained in the one session with the 99 minute target presented for 1.5 seconds. These data taken in acomparison with the data from condition 1 give us a measure of the extent to which warning as to when a target will appear influences the contrast thresholdo Thresholds are presented for each of the three observers below0 Observer Threshold Contrast S.0197 N ~o203 Average.0178 0 These data suggest that a confrast factor of 1.40, is required to compensate for the loss in target detectability introduced when observers are not notified in advance that a target is about to be presented. Tkee session$ were conducted in whicht the observers were given no warning as to when to expect a target and in which the average of the random intervals between targets was 28 seconds. The 99 minute

The University of Michigan ~ Engineering Research Institute 2455L2a-F target was used throughout' b t the duration was varied randomly from trial to trial among.the three possible valueso The three sessions were identical; three were used so that there would be as much data for each of the unknown durations as had been collected for each target condition under condition 1. After all three sessions had been completed,, the data were tallied separately for each duration used. The threshold contrast values for the presentations involving the 1o5 second exposure duration are presented below, Observer Threshold Contrast S.0276 N oO2C3 Average 00203 These contrast thresholds may be compared with those obtained under condition 1 to assess the effect of the absence of warning that a target is to be presented and t-he absence of knowledge of its exposure duration. A contrast factor of l"6 compensates for the restriction of information involved in this experimen tal comparisono Condi:tioon 4 Three sessions were again conducted in which the.observers were given no warning as to when to expect a target and in which the average of the random intervals between targets was 28 seconds. The thrlee target sizes were presented in random order with a fixed and known target exposure duration of lo5 seconds, All targets were used in all three sessionsl After completion of the sessions, the data were tallied separately for the different target sizes. The threshold contrast values for the presentations involving the 99 minute target are presented below o Observer Threshold Contrast S 0258 Average 00190 These contrast thresholds may be compared with those obtained under condition I to assess the effect of the absence of warning that a target is to be presented and the absence of knowledge of its size, A contrast factor of 050 compensaes for the restrict ion of information involved in this e:xp erimental comparison0

The University of Michigan ~ Engineering Research Institute 2455-12-F Condition 5 Nine sessions were conducted in which the observers were given no warning as to when to expect a target and in which the average of the random intervals between targets was 28 seconds. All three target sizes (7.'51t c.99:g tad;,1L:..!tiU e.s;:. and all three exposure durations were used with a completely random order:'.of both variables, A total of nine sessions were required to insure that the data for each target size and duration would equal in number the corresponding data from condition lo After completion of all nine sessions, the data were tallied separately for each target Thre$shold contrasts are presented below, for the 99 minute target presented for a l15 second duration. Observer Threshold Contrast S o0227 N.0182 B,ol43 Average,0184 As before, these thresholds may be compared with those obtained under condition 1 to assess the effect of the absence of warning that a target is to be presented and the absence of knowledge of both its size and duration, A contrast factor of 1.45 compensates for the restriction of information involved in this experimental comparison, It is curious that this factor is actually smaller than the factor for removal o0 knowledge of size alone (1,50) or duration alone (1,60), although perhaps the difference is not statistically significanto These data suggest that the absence of warning that a target is to be presented requires a contrast factor of 1.40 and that the additional removal of knowledge concerning target size, or durations or both size and duration increases the factor to an average value of Lo52 (average of 1.50, 1.60, and 1.45). Thus, the additional contrast factor needed to compensate for removal of knowledge of size or duration or both, when no warning has been provided, is 10o09o Thus far, we have utilized only a small part of our data, having restricted the analysis to cases in which the target subtended 99 minutes and the duration was 1.5 seconds so that direct comparisons could be made with the data obtained in condition 1. We may, however, make additional comparisons among our data by comparing data obtained in conditiors 2, 3, and 4. In all these conditions, there was no warning as to when the target-was to be presented so that we are analyzing possible effects which may exist in addition to the effect of warning. First, we have considerable data on three target sizes and three durations under conditions 2 and 5, (Heretofore we have analyzed only the data for the 99 minute target and the 1.5 second duration,) Combining the data for all nine targets in each case, we obtain the

The University of Michigan ~ Engineering Research Institute 2455-12-F threshold contrasts presented below. Threshold Contrast Observer Condition 2 Condition 5 S.0239 ~0274 N o0198.0203 B o188.135 Average.0188,0204 A comparison between these data evaluates the effect of removing knowledge of both size and duration, with the absence of warning in both cases, A contrast factor of 1.09 compensates for the restriction of information of this sort, (This factor agrees exactly with the contrast factor obtained by averaging instances in which size alone, duration alone, or both size and duration were unknown for the 99 minute target presented at a 1,5 second duration, This agreement is impressive since very few data are common to the two analyzesv:i, It is suggested that, indeed, it makes no difference whether knowledge of size alone, duration alone, or size and duration is removed).. There are two additional comparisons that can be made, We may compare all the data obtained in condition 3 with similar data from condition 5, to evaluate again the effect of removing knowledge of target size when duration gishal ei:,a dy. airnkta:;Thrih'Ol1d contrast values are presented be low. Threshold Contrast Observer Condition 3 Condition S.0297.0251 N.0209.0192 B.0147:0133 Average o0218.0192 These data show the apparently paradoxical result again that the thresh hIold is lower when knowledge of size is removed than when it is present, provided knowledge of duration has been removed. Finally, we may compare all the data obtained in condition 4 with similar data from condition 5, to evaluate again the effect of removing knowledge of target duration when size is already unknown. Threshold contrast values are presented below. Threshold Contrast Observer Condition 4 Condition 5 S.0311.0255 N.0209.0194 B.0141.0139 Average.0220.0196

The University of Michigan ~ Engineering Research Institute 24,55-12-F These data show the same paradoxical result that the threshold is lower when knowledge of duration is removed than when it is present, once knowledge of size has been removed, Since this result has occurred three times in three analyses, it may be a real finding, It is not at all clear what mechanism can be adduced to explain this result, Perhaps the most general conclusion which can be reached is that the removal of knowledge of target size or duration or both has little additional effects when warning as to when the target will be presented is not provided. If we ignore conflicting secondary trends, we may average together the primary data from conditions 2, 3, 4t and 5 and compare the threshold contrasts under these conditions in which warning was removed with the data obtained in condition 1 We conclude that a contrast factor of 1o49 (average of l,O1, 1.50, 1.60 and 1,45) compensates for the absence of warning and for removal of knowledge concerning target size target duration, and both target size and durat ion o Experiment II In this experiment, the background luminance was maintained at 400 foot7lamberts, This experiment was intended to extend the range of values of target size and duration which the observer had to expects in order to evaluate to what extent the conclusions of Experiment I were dependent upon the narrow range of values of these variables which the observer had to expect. In this experiment the following target sizes were used: 2.0, 5o6, 15j7, 44,2, and 92. minutes of arc, The exposure durations used were as follows: 0o3, 0,o6 0o8, and 1,6 seconds, Basically, two experimental conditions were studied, in neither of which the observer was given warning as to when the target would appear. Condition 1 Separate threshold determinations were made for two observers under each of the twenty target conditions during which the observers were informed of the target size and duration, The intervals between successive target presentations were random, with an average length of 28 seconds, as in Experiment I, Continuous curves of size and threshold contrast were plotted for each exposure duration. These curves were used to equalize the visibility of the various targets used in condition 2. Condition 2 A $:ingle experimental session was conducted in which a tremendous variety of target sizes and durations were presented randomly, In all, twenty different target sizes were used at each of the four exposure durations, Each target was presented with an appropriate group of neutral density filters to render it equal in visibility to each 10

The University of Michigan ~ Engineering Research Institute 24I55 12-F other targets based upon the smooth curves developed from the data of condition lo Under these conditions, all the data from all targets may be treated as one group of data, and the session as a whole may be compared with the earlier sessions to evaluate the effect of removal of virtually all information concerning target size and duration. The threshold conrast in the session with targets of unknown size and duration differed by only 3% from the corresponding values when target size and duration were ktnown, This experiment therefore confirms the results of Experiment I in demonstrating that, once warning concerning the time of occurrence of the target is removed, there is no important additional loss in detection due to removing knowledge of target size and duration. These two experiments taken together suggest that a contrast factor of 1.49 is suitable to allow for the deleterious effects due to the absence of warning when a target is to appear, and the elimination of knowledge concerning the target's size and duration.,._ _ _ _ _ _ _ __..11

The University of Michigan ~ Engineering Research Institute 2455-12-F III e EXPERIMENT II The Effect* of Knowledge of Target Location Upon Target Detectability* Ao Introduction It $is almost invariably the case in laboratory studies of visu$al tihresholds that- the observer is givezn complete information as to the location in which the targets will be presented. Obviously, many practical military situations present the observer with targets in unknown locations0 The present experiment was designed to assess the effect upon target detectability of restricting the observer's information as to the location it will occupy. There have been several studies of visual detection thresh'6lds in which the observers were not given complete advance information as to the location in which the target would appear0 For example% some of the writer's Tiffany experiments (Ref. 1) involved the presentation of a target in one'of eight known locations on a circular orbit around a fixation mark. The observers were instructed to scan among the eight locations within a 6 second period during which the target was exposed. The observers moved their eyes around the target orbit in an unknaown manner, picking up the targets at unknown times during their scan, Essentially the same procedure was used by Lamar, Hecht, Hendley, and -Shiaer (Ref, 2), In these experiments, precise information about target location was lacking and the observers were allowed to search and scan for the targets. Since the observers were allowed to move their eyes in the scanning process, it is uncertain to what extent the data reveal the effect of lack of information as to the location of the target, That is, it is not meaningful to compare the detection thresholds obtained in these experiments with thresholds obtained when observers had complete knowledge of tlhe location occupied by the target as an index of the effect of lack of knowledge of target location, In addition to the informational variable, the thresholds for detecting the target in the search case are affected by the fact that the observer often fails to look precisely at one of the target locations lduring the scanning process, It is also apparent that the effective exposure duration is reduced in the scanning psrocess and this will also affect the data obtained in the search caseo It is essential that we identify the elements of the problem involved in searching for targets which will occur in unknown locations, | There is first the effect of information concerning target location *~This experiment was conduacted by Wilson P0o Taannera,. Jr and Margaret.Markerto It was jointly sponsored b1y Plroject MICHIGAN', operatingP under Contract DA 36-0O39-SC52654i between the US, Signml Corps and the University of Michigan, ____ ___ ___ ___ ____-__ 12 __

The University of Michigan * Engineering Research Institute 2455-12-F Se.o There are the additional effects of off-axis viewing and of reduced exposure duration0 "Totalfistic " experiments such as those conducted earlier (Ref.o 1, 2) have the serious limitation that they do not separate these problems and that the data from such experiments cannot have any appreciable generality for application to practical visibility problems. During the Tiffany experiments involving searchl the writer used to emphasize this lack of generality by noting that the results of the experiment would presumably have been seriously affected by a c6hange in the dimension of the orbit around which the target could appear, or a change in the cues used by the observer to identify eadchof the eight possible locations for the targeto Thus, the long-dead Mr. Tiffany could have influenced the results of these experiments by having used a different design in his bowling alley which was used as the experimental roomo Accordingly, the writer decided that the problems involved in searching for targets in unknown locations should be identified and subjected to separate study, Once these aspects of the complex act of search and scanning were understood, actual practical visibility situations could be described by a suitable combination of the various effectso A separate report in this series (Refo 3) contains the results of studies of the effect of targets appearing at various distances off the line-ofsighto A second separate report (Refo 4) contains the results of studies of the effect of reductions in the duration during which a target was visible on the line-of-sight. The present experiment was concerned with the effects of information concerning. the location of the target, and this effect alone o Bo Aparatus and Procedures A complex apparatus was to have been developed which would have permitted us to present targets at virtually any location in a visual field, without the observer's knowledge, Initial stages of the development were completed and the present experiment was conducted even though the apparatus was never brought to completiogn. The apparatus consisted essentially of a projection system mounted on a set of horizontal and vertical ways, By suitable drive mechanisms operating on the two ways, the projector could in principle be positioned at any location within a large area, The entire mobile projecttion syst was mounlfted beEhind a large translucent screen and - was used to project a small increment of light through the screen, This spot appea redt appeared as an increment added to the general luminance of the screen$ produced from a system of diffuse luminaires located in front of the screen, The translucent screen consisted of a sheet of plate glass, with a thin sheet of translucent milk-plastic adhered to tts front surface, The thinnes$ of the plastic sheet tinsured that the spots projected from behind retained sharp edgeso One-half ainch plate glass was used t:o eliminate troublesome ilnter-reflections within the plastic-glas$ sanmdwich o 13

The University of Michigan * Engineering Research Institute 2455-12-F Problems related to the precision of movement of the projector along the double ways proved difficult, as did the problem of eliminating all sound clues to the- location to which the projector was being moved0 These problems were not satisfactorily solved~ However, it proved possible to use a single way to move the projector back and forth horizontally. Thus, in the present experiment's the projector was restricted to one of two possible locations and was made to occupy these in a 9sequence unknown to the observer0 The temporal forced-choice psychophysical variant of the method of constant stimuli was used0 With this procedure, the observer is required to identify in which one of four possible time intervals the target appears, It should be emphasized that in these experiments the observers were always warned precisely when the target presentations would be madeo Successful target detection can be assumed to the extent to.'which the observers can correctly identify the time interval, allowance having been made for chance successes. The allowance for chance success is made from the relation pe 2 (2) 1= ~25 in which p0 = corrected probability; and p raw probability0 All measurements were made with a single target intensity, to simplify equipment requirementso Thus, we did not obtain a frequency-of-seeing curve as tI customary in our use of this psychophysical procedure o The basic procedure was to conduct three related threshold measurements as followso Fiht the target was presented in a known location off the line-of-sighe, for example 89 to the right of fixationr The subject carefully maintained fixation and did not "cheat" by looking toward the target in the known off-axis locationr A run of 50 presentations was made and a value of p0 obtainedo Then., the target was presented in a second known location off the line-of-sight, for example 8 to the left of fixation and a second value of p1 was obtainedo Finally, a series was conducted in which the target could be presented in either one of the two loations$ previously studiedo The mobile optical system was positioned in first one and then the other of the two positions, in acEordance with the requirements of a random program. The observer was informed that the target would appear in one or the other position, and that the probability of its appearance was$eq4aL'ai4hAn two positions in the long run0 The observer carefully maintained fixation midway between the two positions$ and did not attempt to look at either of the two possible target locations0 A run of 5 presentatione was made under these conditions, and a vailuef of p0 obtained0 Under all three csonditions, the observers had merely to identify in which of the four possible time intervals the target appearedL -c.'{';i? 4.dition`;'of'{~nk.owiiiwlocati on, t@here was no requiremenat that the obs$erver be aware of which position the target oc~upied0

The University of Michigan * Engineering Research Institute -2455X l2r Thus, conceivably, the observer could have correctly identified the time interval without ever being aware in which position the target occurredo After a series of three such measurements, the values of p, obtained with the target in the two known locations were averaged0 The value of pv obtained when the target was presented in one or the other position was compared with this value to provide an assessment of the effect of knowledge of target location0 The values of p9 give us information concerning differences in detection probability due to the presence or absence of knowledge concerning the location of the target0 However,, it is more useful to determine the target contrast required to compensate for this effecto Values of the contrast factor needed to compensate for this effect were computed. from the pairs of values of p'0W utilizing data ohthe shape of the frequency-of-seeing curve presented elsewhere by tlh writer (Refo 5)o These data give a value of the slope of the frequency-of-seeing curve measured by a, with respect to the threshold,, specified by Mo A value of a/M:.390 was used for this conversion. In different series of experiments, different pairs of possible target locations were utilized0 In each cash the two locations were separated by equal distances from the line-oftsight, along the horizontal meridian0 In different experiments, the separation between the two locations was varied from 0 25 to 8 degrees$o Differences in the separation between these locations are consideredLto represent differences in the knowledge' available to the observer about target location. C, Results The results$ of these are riesented in Figure 1, The target contrast values are presented in terMs of the contrast factor required to compensate for lack of knowledge concerning target location, As indicated in Section B above, these factors were derived from pairs of values of p9 by reference to average frequency-of-seeing data0 In effect, the contrast factor represents the increase in'target contrast required to restore the value of p9 obtained without knowledge of target location to the value obtained with such knowledge, Contrast factors are presented as a function of the separation between the two possible locations in which the target could appearl It is apparent that the contrast factors exceed unity, and in general increase as the separation is.increased~ (The experimental point for a 2 degree separation between the two possible locations is extremely high for no known reason.) The smooth curve drawn through the rather scattered data has an asymptotic value of 1,310 This means that when the obsesrver must expect a target to appear in either of two possible locations, and when these locab tionrs are separated by at least 4 degrees, the target contrast must be increased by a factor of 1,31 to compensate for the reduction in information made available to the observer, It must be emphasized that this factor does not take account

The University of Michigan ~ Engineering Research Institute 2455-12-F of the contrast increase needed to detect off-axis targets, Our procedure compares situations in both of which the targets appeared off-axis to the same extent. but in which the information made available o e to the observer about target location was the experimental variable. It is not clear to what extent these data can be used to represent realistic visibility situations in which the target mus$t be expected in many possible locations over a fairly large area of the visual field. It might seem as though two possible positions represent only a very small uncertainty as to target location, in comparison with that encountered in most practical situations, HIowever, it may be that once specific information as to the precise location of a target is removed, further reduction in location information may have very little effect upon target detectability. Obviously, further experiments along these lines should be conductesd, in which increasing uncertainty is introduced with respect to the location to be occupied by the target. These experiments might well be guided by considering the visual system as being "scanned" at higher neural centers by a mechanism with a limited area of "attention"~ In such a constructs the probability of detection will be the probability that the scanner will intercept the target while it is still exposed. Other theoretical models which should be considered assume that the visual neural system has a decision process in which signals (targets) must be differentiated from noise (no targets)o In such a system, certainty with respect to target location can reduce the se in the system by eliminating from consideration the neural events occurring at other locations in t0e sys t em,

The University of Michigan ~ Engineering Research Institute 2455 12a-F IV o EXPERIMENT III The Effect of Observer Criterion Upon Target Detectability A0 Introduction It has been standard practice in these laboratories for some years now to measure detection thresholds by the temporal forced-choice method of constant stimulio This method is charaterized by the requiremenit that the observers indicate their ability to detect the presence of a target by correctly identifying the time interval out of four possible itervals o Extensive studies reported elsewhere (Refo 6) demonstrate the excellent reliability and validity of this method, which make its choice a happy one for the kind of extended program of research conducted by these laboratories$ The forced-choice method requires that the observers select a temporal interval even when they have no confidence that they can detect the presence of a target, After practice, observers universally become able to identify with a high degree of accuracy targets which were originally "invisible" to them0 Studies reported elsewhere (Ref0 6) suggest that observers not only learn to detect targets of very low contrast with the forcedchoice mfethod but also learn to estimate how accurate their forcedchoices are0 The observers apparently come to realize that very dim and indistinct experiences do represent the presence of real targets, provided these experiences have prope'r characteristics with respect to time and place of occurrence and sharpness of onset and offset0 This by no means implies that the observers can utilize what they have learned in practical visibility situations. Under realistic conditions, targets do not oblige by appearing in known locations at known times and theiy do not have contrsolled rates of appearance and disappearance. These consi$derations suggest that what is learned in the forced-choice laboratory situation may be relatively useless to an observer in a practical situation0 It may be that there is some generality in learning to detect dim targets, but this has yet to be established0 The problem, therefore, is to assess the extent to which the criterion used by trained observers utilizing the forced-choice procedure differs from the kind of criterion which will be used by military observers0 An evaluation of this difference will permit us to define a contrast factor to compensate for this difference when we apply laboratory detection data to practical visibility situations in the field0 The data to be reported here were collected under a program $ponsored by the Office of Naval Research and have been reported elsewhere (Refo 6)o These data have been analyzed in a new way for the present purpose and hence the conclusion is new even if the data

The University of Michigan * Engineering Research Institute 2455-12-F are old. Under these circumstances~ it will suffice to provide only a brief description- of the apparatus and procedures used to collect the original data, The first phase of the present analysis involves comparing detection probabilities with the forced-choice method with those obtained with the "yes - no" method, in which the observers indicate by "yes" or ".no" whether or not they have detected the presence of the targeto For this purpose, data were utilized which were obtained on 4 unusually experienced observers. These observers utilized the forced-lchoice and the yes - no procedures alternately from day to day in a series of daily experiments extending over more than 10 months0 The target and background conditions were maintained constant during this period so that these observers had a stalggeringLly large amount of experience with this one detection situation0 The general luminance was 4o71 foot-lambertso The target subtended 1805 minutes of are, and was presented always 7 degrees from the line-of-sight for about 0.072 seconds0o These observers had every opportunity to develop confidence in their forced-choice respons@es and to attach t:he verbal symbol "yes" to the experiences of dBim and vague targets Ewhich led to corrdect forcedchoice responses. Under these conditions, the difference between the probabilities of detection obtained with the forced-choice and the yes no methods should be minimal The second phase of the present analysis i7nvolves an eviluation of the extent to which the probabilities of detection with the yes no procedure improve with practice. For this purpose, we have analyzed data obtained under somewhat different experimental conditions on an entirely different group of observers0 A group of 70 observers was used in several series of experiments in which different methods of training were evaluated0 In all cases$, a point source target was employed in a known location on the line-of-sight. The exposure duration was 15 secondso Background luminance varied in the different experiments from 17o9 to 18.7 foot-lambertso Each observer was introduced to the experimental situation with a set of instructions intended to elicit what might be called a "commonsense criterion" of seeing. The observers were told: "We are going to turn on a light from time to time0 If you see a light, say'yes'0 If you don't, say'no". (We did not inform the observers that we were presentinLg blank trials to evaluate their criteria.o ) If the observers asked, "How will I know?"t, we told them: "Oh, you'll know when you try it." These instructions were intended to keep the o>bservers as naive as possible and to prevent t~hem from devrelotping a laboratory frame of reference Sublsequently, these observers were divided into groups who were given different instructions and training0 Some of thEese led the observers to iprove their detection probabiities a great8 deal 18

The University of Michigan ~ Engineering Research Institute 24~55 l2o F more than otherso, In order to assess the relation between the c nommoense criterion and the usual laboratory criterion, we have compared the results of all the observers in the initial experimental se$sion with results$ frXom the observers who u-tiLized the most effective regimes of instructions and trainingo Large enough groups of equivalent observers are involved so that this c.mp arison is not seriously affected by sampling differences, The final phase of the present analysis involves combining the results from the two phaseso It is assumed that the yes - no criteria adopted by the 4 highly practiced observers were at least as effective as the criteria adopted by the observers from the large group who were found to have experienced the most effective regime of instructions and trainingo (It is not possible to test this assumption because of the differences in the experimental situations involved in the two experimental phaseso) On this basis, the data from the second phase can be combined with the data from the first phase, to reveal the extent to which the commonsense criterion results in fewer detections than the laboratory criterion used in the forcedchoice procedure. Co Results The data from the two phases are presented in Figure 20 The frequency-of-seeing curve plotted the farthest to the left represents the results obtained by the 4 highly practiced observers with the forced-choice pr.ocedureo- The relative contrast is set equal to 1 for a detection probability of 0,50, The frequency-of-seeing curve is constructed with a/M: o390, which was found to be the average of many detection experiments, as has been reported elsewhere (Refo 5)o The frequency-of-seeing curve second from the left, which represents the left border of the hatched area, has been constructed with a contrast factor of 1.20o That is, this curve is constructed with a/N M.390 as with the first curves but the contrast at which the detection probability 0o50 is set at lo:20 The two curves farthest to the left in tae figure, represent the data obtained by the 4 highly practiced observers with the yes - no mefthod in comparison with their data obtained with the forced-choice methodo The frequency-of-$eeing curve farthest to the right, which represents the right border of the hatched areas has been constructed with a contrast factor of 2o40, That is, this curve is constructed with a/NM 390 as with the first curves but the contrast at which the detection probability - 0o'5Q is set at 20400 This curve has been constructed to represent the fact that the commonsense criterion resul~ted in threshold co9ntrnast values for the 70 observers' 2Os0 times the values$ obtained by the observers from the group who were given the most effective instructions and training0 Thus, it is assumed that the com ese ci trion resltos in t hreshold contl rasCtt values at least 200 x lo20 = 20 times h alues eied by the 4 highly practiced observers, utilizi~ the forcedschoice procedureo 19

The University of Michigan ~ Engineering Research Institute 2455-12-F The hatched area is considered to represent a range within which different criteria with the yes - no method may occurs depenhding upon the extent to which observers have learned to be aware of the most dim and vague cues available to them in the laboratory. It is not entirely apparent what factor should be utilized in converting laboratory data for use in practical visibility problems. Since the field observer rarely has opportunity to use vague cues which occur in the laboratory, the 2.4o factor is probably most reasonable0 Use of this factor does not involve an extreme allowance for the effect of observer criterinon since our process of combining the data from the two experimental phases was conservative,

The University of Michigan ~ Engineering Research Institute 245512 -F.V0 EXPEBRIENT IV The Effect 6f Frequency of Target Occurrence Upon Target Detectability* A Introduction It is customary in. laboratory investigations of visual thresP holds. to present targets frequently, and to warn the observer before each presentation is made0 Military targets usually occur without warning and they seldom occur with the frequency of laboratory test targetso The present experiment was designed to assess the effect of the frequency of targets upon their detectability, when the targets were presented without warning0 There has been considerable interest among the English psychologists in the last few years in what they call vigilance (Refo 7)0 By vigilance is meant the'alertness of the observer, measured by his sensitivity to stimulus presentationso Results of several experLimeints suggest that vigilance is reduced by-,reducing the frequency of stimulus presentationso These results have only been obtained in comparatively complex sensory-motor tasks$ We have wondered to what extent reducing the frequency of target presentations will influence the threshold of visual detection0 Presumably the observer could easily maintain his vigilance, if he were always warned before a target was to be presented. Therefore, we have studied the influence of the frequency of target presentations upon visual detection in a situation in which the observer was not warned before the target was to be presented. We have measured the probability of visual detection with each of two frequencies of unwarned target presentationSo In each case, the observer was required to detect the presence of a point source target which was added to a screen of moderately high luminance, B, Apparatus and Procedures Four simple light boxes were cons~tructed to provide fields of uniform luminance, Wooden cubes were constructed, measuring'24 inches on a sideo The observers viewed the back wall of each cube through an open aperture on the front wall. The back wall consisted of a thin milk-plastic screen which was uniformly illuminated by ated y frosted lamps mounted inside the cube, but invisible to the observers0 A metal plate was mounted flush with the rear of the plastic screens witlh a small hole which could be illuminated by a projection lamp mounted behintd the screenmo When tihe Laperture was ill fuminatedg a small spot of ight was added to the luminanc$lE of the screen~ The intensity of the luminanace increment- produced when the spot was exposed was governed *~This experiment was conducted by Celeste Mo ECrossnman

The University of Michigan ~ Engineering Research Institute 2455- 122 F by Wratten neutral ldensity filers interposed Ibetween the lamp and the screen0 The target could be occluded with an -opaque shutter, or presented by removal 6f the shuttero In the present experiments, the shutter was operated manuallyo It was always opened for 2 seconds, the timing being controlled ty the experimenter with the aid of a stop watch. The luminanee of the plastic screens in the four boxes varied from 6o67 to 7.29 footLambertso Variations in these lumrinance values from day to day were neg igibLe throughout the experiment0 Each observer sat at a distance of eight feet from the screen0 He was required to view the screen continuously and to report whenever a target was addedo In order to control accommodation and orientation, four black dots in the form of a diamond were painted onto the center of the plastic screen0 The angular subtense of the target and of the black dots was 1 minute of arc0 Each of the black dots was located 10 minutes of arc away from the central location occupied by the targeto Each observer indicated that he detected the presence of a target by depressing a doorbell buttono The button activated a neon glow iamp located in the anteroom behind the light-box where the experimenter sato Four light-boxes were assembled in a row so that four observers Icould be used simultaneously. It was essential that multiple observers be used, in order that reasonably efficient use could be made of the experimenter's timeo Black drapes were hung from ceiling to floo; between the light-sboxes so that the observers were unable to see each other, or to see more than their own light-box. The experimental room in which the apparatus was located was provided with an air-conditioning unit0 This unit was operated at all timeso The steady noise it made was more than adequate to mask any sounds made by the observers or the experimenter0 The basic experimental plan was to compare the probability of detection of unwarned target presentations under two conditions, designed to represent extremes with respect to the frequency of target presentationso "Frequent presentation" was defined by one target presentation every thirty seconds, on the averageo "Infrequent presentation" was defined by one or two presentations every twenty minuvtes, on the average. The observers were firsT given several experimental sessions in which the presentations were frequento They were then given a number of sessions in which the presentations were infrequento The first -condition was thenl repeated in order to provide an approximgate control for temporal effects~ such as learning0 Thie observers were obtained from among University students desiring parttime employment, and were paid for the timae devotedL to the experiment0.. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 2 2

The University of Michigan ". -. Engineering Research Institute 24%55,-12-F It was expected that the observers would indicate that they had detected the presence of a target when a target was not presented an appreciable percentage of the time0 The number of such spurious responses was ~, small throEughout the entire experiment, however that they may be safely gmnored in the analysis of the data. Twelve observers were scheduled to be used in three groups of four each, These observers observed for five experimental sessions of two hours each under conditions of frequent presentationo Each session was divided into five sub-sessions of twenty minutes each, with a rest period of five minutes following each sub-session0 In each sub-session, forty presentations were made without warning, at randomly selected timeso The targetIs were presented simultaneously to all four observers since any other arrangement would have been impossible for the experimenter to' manage Equal numbers of each Of five target intensities were presented in random order in each experimental sessiono The observers were then instructed as follows: "From now on. you will see fewer lights0 Remain as attentive as possible and try to see as manyy s possible0" They were each given a total of sixteen e@xperimental sessions under conditions of infrequent presentation0 Each subsession was desigPned to have either zeros one, twos or four detections by the observer0 Successive sub-sessions varied in the number of detections which it was desired that the observers make, within these limits. The number of sub-sessions in which each number of detections was to occur was the same~ and the schedule was arranged so that a sub-session with zero detections followed a sub-session with one detection as often a* it followed a sub-session with two detections, and so ono In these experimental s$essions, the experimenter employed a different schedule for each of the four observers working at the same time, never presenting targets to two observers at the same timeo In a given sub-session, the experimenter kept presenting targets until each observer had obtained his quota of detections for that subsession, This arrangement was impossible to adhere to precisely, but only minor exceptions were necessary0 In order to obtain the desired number of detections, the experimenter found that she often had to repeat the presentation of targets in fairly rapid succession. Two restrictions were.made on the way in which the targets could be presented. First, target presentatfions could never Be closer together than 15 seconds0 Second, no presentations were permitted during the last 2 minutes of each sub-sessiono Five t;arget intensities were presented in random order as before0'The intensities were selected on the basis of the results of the earlier experimental sessions, with an an allowance for the extent of the reduction itn detections which was expected to occur when target presentations1: were'reduced in frequencyo The allowances made for the expected effect turned out to be reasonably accuratee During t~he last few experimental sessionss, two of the observers gave evidence that they had become aware that the experimenter repeated 23

The University of Michigan * Engineering Research Institute 2455-12-F targets which were not detected the first time they were presented~ One observer reported that, if he thought he saw something, he waited until it was presented again-:.t be sure of his response, The:other observer told the experimenter he' thought she repeated target presentations every 15 seconds, some of the timeo As noted above~ this interval was exactly the shortest allowable interval separating successive target presentations. The demands on the experimenter to keep four:observers to their own prearranged schedules were so exacting that she was often forced to use this short a separation between successive presentations0 The fact that one observer surely was aware of the mode of presentation, and a second probably was aware of the mode of presentation, suggested that the data obtained in these sessions should be discardejo At this point in the study, one of the twelve observers quit for personal reasons and the two observers who had suspected that targets were: being presented repeatedly, if missed, were dropped from the experiment. The remaining observers were next told that the targets would be presented more frequently again, as in the original experimentso Two ~sessions were given each observer, in which forty target presentations were made in each sub-session, Following these sessions, the observers were told that the targets would be presented less often again0 The succeeding fourteen sessions involved infrequent target presentations$ in accordance with a revised plan0 Instead of specifying that the observer should obtain a certain number of detections in a given sub-session, the number of target presentations to be given in each sub-session was set in advance and rigidly adhered to0 The fourteen experimental sessions contained seventy sub-sessions of twenty minutes each0 Forty of these sub-sessions contained exactly one target presentation; twenty contained two presentations; and ten contained four presentations0 All the presentations were made at one target intensity in order to concentrate the data to the maximum possible extent0 The order of sessions involving ones two, and four target presentations was randomized, The times during the sub-sessions at which the target presentations were made was controlled to the extent that an equal number of presentations was made on the average in each eighth of the sub-session. The target intensity at which all the presentations were made was one of the five intensities which the observer had used during the sessions with frequent presentations conducted just prior to the sessions. with infrequent target presentations0 When all fourteen sessions had been completed, the observers were again given two experimental sessions in which frequent target presentations were made0 The same five target intensities were employed as in the sessions with frequent presentations which preceeded the sessions with infrequent presentations0 Co Results All the data obtained in the -fourteen sessions in which targets were presented infrequently were combined0 T1he data obtained with the target intensity in the sessions involving frequent presentations, con'ducted just before and just after the sessions with infrequent

The University of Michigan ~ Engineering Research Institute 2455-12-F presentations,, were also combined. The resulting proportions, P. and the corresponding values of N, the number of target presentations made, are presented in Table I for each of nine observerse Values of wp were computed by means of Bernouilli's theorem, and the significance of the difference between the pair of proportions obtained by each observer was evaluated by the critical ratio test. Values of P(CR) represent the probability that a difference in either direction as large or larger than that obtained could have occurred by chance, We see that eight of the nine observers show highly significant differences in P. dependent.upon the frequency of target presentations. Six observers show the expected loss in detection probability for infrequent presentations, whereas two observers show a significant' gain: in detection probability. The "average effect" may be judged by the average proportions obtained by all observers under the two conditions. There is a comparatively small loss in detection probability, on the average. It will be of interest to express the average effect of the frequency of target occurrence in terms of the contrast increase required to compensate for a low frequency of occurrence. Data on the form of the psychophysical curve obtained in visual detection experiments presented elsewhere (Ref. 5) may be used to compute a suitable contrast factor. Utilizing a value of c/M =.390 as before, we obtain a contrast factpr of 1.19. This means that a 19% increase in target contrast will compensate for the effect of reducing the frequency of target occurrence as was done in the present experiment. Since the frequency reduction was quite extremes use of a 1.19 contrast factor to allow for this factor in practical military visibility problems seems quite safe.

The University of Michigan ~ Engineering Research Institute 2455 12-F REFERENCES Blackwell, Ho R, "Contrast Thresholds of the Human Eye". Jo _.t Soc.o Amer., ", 624-643 (1946), 2 Lamar, E. S., Hecht, S., Hendley, C. D,, and Shlaer, So "Size, ShapeP and Contrast in Detection of Targets by Daylight Vision: I Data and Analytical Description", J. oto Soc_ Amer., 7, 531 545 (1947)G 3 Blackwell, H, R, and Moldauer, A. B "'Detection Thresholds for Point Sources in the Near Periphery". University of Michigan, Engineering Research Institute Report 2455-14-F (in press), 4 Blackwell9 H. Ro and McCready, D, W,, Jr. "Foveal Contrast Thresholds for Various Durations of Single Pulses", University of Michigan, Engineering Research Institute Report 2455-13-F (in press). 5 Blackwell, H, R, "Brightness Discrimination Data for the Specification of Quantity of Illumination". Illum. En,. N.Y., 4, 602-609 (1952), B6 Blackwell, Ho R, Psy;hophy$ical Thlresholds: Experimental Studies of Methods of Measurement. University of Michigan, Engineering Research Bulletin No, 36, 227 p (1953), 7 Mackworth, N, H.. "The Breakdown of Vigilance During Prolonged Visual Search", Quart, _J Exp, Psychol. 1, 6-22 (1948),

TABLE I Detection Probabilities for Frequent and Infrequent Target Presentations 1 Observer Frequent presentations Infrequent presentations Significance of difference Direction o P N p N Critical ratio P(CR) of effect 1.810 100.438 114 6.11 <001 Loss 2.800 0.426 115 4,77 <.001 Loss < 3.891 120.478 115 7.56 <.001 Loss 4 o741 116 *393 117 5.74 <001 Loss 5.629 116.828 105 3.44 <.001 Gain 6.500 80.792 120 4.34 <.,001 Gain "I 7.736 110.325 120 6,84 <;001 Loss 8.692 120.687 115 0.08.936 None 9.760 75.475 120 4.24 00 Loss Averages.729.537 Report 2455-12-F - - l~~~~~~

GA 10(7 1. 0 1 2 3 4 5 6 7 6 Separation between Possible Locations (degrees) Fig. 1, Effect of absence of knowledge of location. GA 1133 o 1.0 2.0 3.0 4.0 RELATIVE CONTRAS T Fig. 2. Effect of observer criterion (see text). 28

3 9015 02523 0445 THE UNIVERSITY OF MICHIGAN DATE DUE