Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
SPATIAL AND TEMPORAL FACTORS AFFECTING NIGHT VISION
INTRODUCTI ON Kenneth R. Alexander The spatial characteristics of vision include visual acuity, which is the ability to discriminate the fine detail of the visual environ- ment, contrast sensitivity, which is the capacity to detect spatial variations in light levels, and the perception of objects and their relationships in space. Temporal processing refers to the discrimina- tion of visual events that vary in time, including the detection of brief flashes, motion, and flicker. In the visual tasks involved in flying, such as reading cockpit instruments or judging terrain during cross-country navigation, the ability of a pilot to assess the spatial and temporal aspects of visual information accurately is critical for optimum performance. However, the capacity of the visual system to process spatial and temporal information varies considerably with illumination level. In a well-lit environment, vision at night may be cone-mediated, as it is during the day. When the illumination level is reduced, both cone and rod systems may govern visual sensitivity, while the roe system may be the primary determinant of visual sensitivity at low light levels. The cone and rod systems differ substantially in their response to spa- tial and temporal visual information. Typically, the cone system pro- vides much finer spatial and temporal resolution than the rod system. Therefore, reliance on rod-mediated vision under dim illumination con- ditions can result in a considerable decrement in visual performance. Even if vision is cone-mediated, however, the level of illumination can have a significant effect on visual tasks. For example, the visual acuity for letters on an eye chart drops from 20/20 (normal) to 20/100 when the chart illumination is decreased by only a factor of 100 (the range of illumination to which the visual system is sensitive is 10,000,000,000 to 10~. Results such as this imply that there is a marked functional reorganization of the visual system that occurs with reduced illumination levels. The papers in this section pro- vide detailed information about the changes in spatial and temporal processing that occur as a result of reduced illumination. The following remarks briefly summarize the issues discussed in these reports. Under daytime illumination conditions, the fovea is the most sen- sitive part of the retina and is used for fixation. However, with reduced illumination, the fovea may no longer be the most sensitive 143
144 retinal region, and fixation may shift to a nonfoveal location. To optimize visual performance, it is important to determine the factors that govern fixation when light levels decrease. In the first report, Harold Bedell discusses the sensitivity acuity and directionalization properties of various parts of the visual field and their relationship to eye fixation under conditions of reduced illumination. Also consi- dered are the potential deleterious effects of optical blur, pupil size, and accommodation on visual performance under reduced illumination con- ditions. The detection and identification of visual targets depend on more than good visual acuity, as demonstrated by a number of recent studies. Lewis Harvey presents a relatively new method for assessing visual per- formance, in which targets are described in terms of their spatial fre- quency characteristics. The detection and identification of these tar- gets are derived from the contrast sensitivity function. As discussed by Dr. Harvey, this method has proved to be more useful in predicting the visual performance of pilots than the traditional measurement of visual acuity. An extension of this technique to the rod system may provide a successful way of predicting visual performance under condi- tions of reduced illumination. The ability of a pilot to perceive the layout of visual space and its relationship to himself or herself is critical for safe navigation. The visual cues that are available for accurate space perception change dramatically with illumination. Ralph Haber details the visual task demands required of pilots in the perception of visual space during both day and night flying. He also discusses the role of individual differences in night vision and the possibility of improved night vision through training. The ability to detect visual targets at night is strongly influ- enced by the presence of extraneous sources of illumination, termed disability glare. The Blackwells note in their paper that disability glare degrades visual performance by reducing the contrast of the reti- nal image of the target through light scatter. Dr. Blackwell provides a quantitative evaluation of the factors that influence disability glare, such as glare luminance, pupil size, and observer age. He also discusses an instrument designed to measure disability glare objectively and a device that measures the effect of disability glare on visual per- formance. The rod and cone systems in human vision do not function indepen- dently. The response of one system can markedly influence the sensi- tivity of the other, often to the detriment of visual performance. Donald MacLeod and Andrew Stockman point out the ways in which rod and cone signals can interfere with each other in such visual tasks as the detection of flicker. He also presents evidence that the sensitivity of the rod system to temporally varying information is substantially greater than had been previously estimated. Patients with "low vision" or severely impaired visual function can exhibit many of the visual problems that normal observers experience under conditions of reduced illumination. For example, reading is it paired in normal observers if the illumination level is too low, just as it is in patients with a loss of foveal function. Ian Bailey
145 discusses the problems of mobility exper fenced by low-vision patients and evaluates the assistance offered by low-vision aids. Methods of improving vision in such patients may be helpful as well in aiding night vision in pilots.
ECCENTRIC REGARD, TASK, AND OPTICAL BLUR AS FACTORS INFLUENCING VISUAL ACUITY AT LOW LUMINANCES Harold E. Bedell It is readily appreciated from casual experience and has been well documented experimentally that visual acuity worsens as the level of illumination is reduced (Koenig, 1897, cited in van Helmholtz, 1925a; Schlaer, 1937~. This worsening of acuity is most pronounced at the fovea, where under photopic illumination acuity is best and under sco- topic illumination the eye is totally insensitive. One contribution, then, to the reduction of acuity under scotopic illumination is the necessity of viewing the target with a peripheral retinal locus that is inherently less acute than the fovea. The shift of the target from foveal to nonfoveal regard is not the entire story, however, as can be seen in Figure 1, in which data from Mandelbaum and Sloan (1947) are replottea on a logarithmic acuity axis. A logarithmic scale for acuity is used because on such a scale response variability remains constant, despite changes of the acuity threshold due to the target's retinal eccentricity (Westheimer, 1979) , luminance (Simpson et al., 1981), or blur (Prince and Fry, 1956; Homer et al., 1985~. Figure 1 shows that between 4 and 8 degrees from the fovea, re- duction of target luminance front about 3 to 0.01 cd/m2 results in a worsening of visual acuity for Landolt ring targets of approximately 2.5-fold (0.4 log units). Further reduction of the target luminance by a factor of 10 (to 0.001 cd/m2) produces about another twofold loss of acuity at the same retinal eccentricities. These results, which are confirmed by the data of other studies in which similar (Sloan, 1968) and different (Aulhorn, 1964; Kerr, 1971; Koenderink, 1978) types of acuity targets were used, indicate that, as in the fovea, acuity in the near peripheral retina also depends substantially on the level of illu- ~r.inat ion. COMPARI SON WI TH OTHER TASKS In contrast to visual acuity, the precision with which targets can be directionalized has been found to be much less dependent on the Supported in part by research grants EY 05068 (to Harold E. Bedell) and EY 03694 (to Merton C. Plot) from the National Eye Institute. 146
147 2.0 .8 .6 .4 1.2 al: ~1.0 o J .8 .6 .4 .2 o o.oo 1 3 cd~m2 o.o 1 3 cd/m2 0 3.1 8 cd/m2 ~ /N ~ -of- _ o _¢ it, ~- ' ~ ~, . __4~ - _ - ~ _ -I - ._ _ _-~ r I , , , , , , , , 1 o 2 4 6 8 10 ECCENTRICITY (DEG) FIGURE 1 Extrafoveal visual acuity data (MAR = minimum angle of reso- lution) for Landolt ring targets at different luminances are replotted from the data of Mandelbaum and Sloan (1947) on a logarithmic acuity axis. level of illumination. About 30 years ago, Leibowitz and coworkers (1955a,b) reported that observers could discriminate the meridian in the field of a near peripheral (ca. 4 degrees from the fovea) target with essentially equal precision whether the target was at its light
148 detection threshold or at much higher luminances. Recently, these findings have been extended (Bedell et al., 1985) by showing that observers' precision in specifying the direction of a target along a meridian of the field (the horizontal) is similarly insensitive to reductions of target luminance to near-threshold values. In our study (Bedell et al., 1985) we assessed the precision of directionalization with a variant of the classical monocular spatial partitioning task (von Helmholtz, 1925; Ogle, 1932~. Observers successively partitioned an 8.3-degree space in the right field into a series of perceptually equal fractional spaces and then matched the direction of each parti- tioning target in the opposite (left) half field. The targets were luminous vertical lines, ranging in luminance from 40 to 0.04 cd/m2, controlled by neutral density filters. For one subject, luminance of the targets was further reduced with added Polaroid filtering to nearly the detection threshold. Neither the precision of directionalization (measured as the standard deviation of the psychometric functions fitted to observers' partitioning judgments) nor the observers' small idiosyn- cratic constant errors depended significantly on the luminance of the targets. Together with the results of Leibowitz et al. (1955a,b), results of our study (Bedell et al., 1985) indicate that the specifi- cation of where a target is in the visual field is essentially inde- pendent of its luminance. ' Figure 2 shows preliminary data (for the author's left eye) on the effect of luminance on another directionalization task in which a Vernier-type stimulus is viewed extrafoveally. The task (diagramed In the inset to Figure 2) is to judge whether the small vertical line (3.S by 30 min) is to the left or the right of an imaginary axis that con- nect~ the tips of the two opposed triangles, which are separated by 1.5 degrees. The triangular reference targets are presented continuously, and the line is flashed for 250 ms. Fixation is midway between two vertical lines (each 1 degree long and separated by 1.5 degrees) to the left of the triangles, a stimulus adequate to promote quite steady fixation (Rattle, 1969; Sansbury et al., 1973~. The data plotted in Figure 2 represent the horizontal distance, in log minutes of arc, that the line must be displaced from its locus of subjective alignment (where right and left responses occur equally often) to increase the probabil- ity of a right response by 1 standard deviation (from 50 to 81 percent). In the range of eccentricities examined, there was very little effect on threshold of reducing the luminance of the display (achieved by inter- posing neutral-density filters) from 40 to 0.4 cd/m2. When luminance was further reduced to 0.01 cd/m2, thresholds' were elevates noticeably at 1 and 2 degrees, but only slightly at greater eccentricities. In fact, the substantially elevated line-displacement thresholds at 1 and 2 degrees when the display luminance was 0.01 cd/m2 are somewhat arti- ficial in that the observer guessed the direction of the line's offset on many presentations at 2 degrees and most of the presentations at 1 degree when the line was not detected at all. When the luminance of the flashed line was adjusted to be near threshold (detected about 80 percent of the time) and the observer judged the liners direction with respect to the triangles only when detection occurred, line-displacement thresholds at eccentricities of 1 and 2 degrees (T's in Figure 2) were
149 2.0 ~ 1.8 - E 1.6 O 1-4 ~ :~ in CC 1.2 _ :~ _ A_ 1.0 _ He E.0.8_ a: CL 0.6- cn LJ 0 4- z - 0.2 o 0.0 T 1 en' An' V · 0.01 cd/m2 0.4 cd/m2 0 40 Cd/m2 /,~ W/~- _ ?~' ? /,? '/~ /; A 2 6 ECCENTRICITY (BEG) 8 FIGURE 2 Thresholds for discriminating displacement of a small vertical line with respect to the axis of a flanking reference target (see the inset) are plotted for the author's left eye. Targets were presented in the nasal visual field at the indicated luminances. Vertical bars on data points represent l standard error. The T's show line displace- ment thresholds when luminance of the l, ne was near detection threshold at that eccentr ic ity in the f isle it deg ree, 0 . 06 cd/m2; 2 deg rees, 0.014 cd/m2; 3 degrees, 0. 005 cd/m ; ~and 8 degrees, 0.004 cd/m .
150 about 0.4 log units poorer than at the highest luminance. At greater eccentricities, line-displacement threshholds were elevated by a smaller amount--only about 0.2 log units--when the line's luminance was at nearly the detection threshold. To compare these results with the other data in Figure 2, it is useful to note that detection thresholds for the line ranged from C.06 cd/m2 at 1 degree to 0.004 cd/m2 at 5 and 8 degrees. Results at eccentricities beyond 2 or 3 degrees are reasonably com- parable to those reported previously for judging a target's meridian (Leibowitz et al., 1955a,b) or its direction along a meridian (Bedell et al., 1985~. Reducing target luminance f ram moderate photopic levels to nearly threshold results in only a small impairment (ca. O. 2 log units, or 60 percent) of relative directionalization thresholds. Closer to the fovea, Vernier thresholds increase about 2.5-fold (0.4 log units) as the target's luminance is reduced to just detectability, a threshold increase similar to those reported for Vernier acuity (Baker, 1949; Berry et al., 1950; Leibowitz, 1955; Westheimer and McKee, 1977) and spatial bisection (Klein and Levi, 1985) at the fovea. Whether or not important differences exist between directionalization tasks is as yet unclear; however, the picture that emerges is that reduction of target luminance has a proportionately smaller effect on the precision of directionalization outside the central few degrees than in the immediate macular area. However, even in the macula, a reduction in target lumi- nance has virtually no effect on directionalization until the luminance falls to within 1 log unit or less of the detection threshold. In contrast, recall that Landolt ring acuity at retinal eccentrici- ties of 4 to 8 degrees decreases more substantially and regularly as target luminance is reduced (Figure 1~. Incidentally, a comparison of the data in F inures 1 and 2 reveals that the Vernier thresholds between 4 and 8 deg rees are similar in absolute magnitude to the acuity thresh- olds for Landolt rings when luminance is 0.01 cd/m2. A likely reason for the relatively poor Vernier thresholds in Figure 2 is that such thresholds are highly dependent on the spatial separation between targets (Sullivan et al., 1972; Enoch and Williams, 1983; Beck and Halloran, 1985~. A related finding is that the separation between the targets at which threshold is optimal increases regularly with eccen- tricity in the field (westheimer, 1982; Yap et al., 19851. Because of the fairly wide separation between targets, thresholds obtained with the stimulus pictured in Figure 2 are not optimal r being on the order of 1 min arc, even when this stimulus is presented at high luminances near the fovea. It is clear that line-displacement thresholds would be improved if the spacing between targets were smaller, and although as yet untested, it is probable that comparable effects on these thresholds would be obtained as luminance is reduced.
151 POS SI BLE INFLUENCE S ON THE ECCENT RI C REGARD OF SCOTOPIC TARGETS The limits of acuity and directionalization in the near peripheral retina are of interest at low levels of illumination because at these low light levels the fovea becomes insensitive and stimuli must be re- garded extrafoveally. An assumption often made when measuring scotopic visual functioning is that observers will use the most sensitive part of the retina for the task at hand if targets are available for a suffi- cient duration and if observers are instructed or trained beforehand to use nonfoveal vision (Schlaer, 1937; Pirenne et al., 1957; Kinney, 1968~. However, a complicating factor is that the retinal locus with the best sensitivity is likely to depend on the task. For example, it is well established that the detection of scotopically illuminated targets is best at an eccentricity of 15 to 20 degrees from the fovea, where the density of scotopic receptors is greatest (Ten Doesschate, 1949; Sloan, 1950~. Scotopic visual acuity, on the other hand, peaks closer to the fovea; Mandelbaum and Sloan (1947) report that, for a range of scotopic luminances, the best acuity is found between 4 and 8 degrees from the fovea (Figure 11. Directionalization thresholds have not yet been investigated systematically in the periphery at low mesopic and scotopic luminance s. However, because directionalization falls off faster with eccentricity than does resolution at photopic levels (Westheimer, 1982; Levi et al., 1985) and, as discussed above, is less affected by luminance, it would not be surprising if at scotopic lumi- nance levels the best directi~onalization thresholds were found closer to the fovea than the peak for resolution. How observers identify and maintain the nonfoveal locus that they use when freely viewing targets of low luminance has not been addressed adequately. In Figure 3, two observers' monocular visual acuity for scotopically illuminated targets is shown as a function of retinal eccentricity. These data, collected in collaborat~on with Linda van Schelt of the Southern California College of Optometry, are for illu- minated Landolt rings (0.004 cd/m2) presented in a dark field. The observer whose data are shown on the right is about a factor of two more sensitive at each eccentricity, which can be ascribed to this observer's much greater practice in making psychophysical aiscrimina- tions in the peripheral field (Low, 1951; Johnson and Leibowitz, 1979~. The arrows in Figure 3 mark the eccentricity to which each observer adjusted a barely visible photopic fixation point to suc~ectively maximize the discriminability of targets somewhat larger than those corresponding to the best measured acuity. Targets were presented continuously both for the measurement of acuity thresholds (50 per- cent, corrected for guessing) and for adjustment of the fixation target. Note that the experienced observer set the fixation mark such that the targets were at an eccentricity slightly greater than that at which the best acuity was found, although the difference in acuity between the two loci is trivial. The second, less experienced observer on two separate occasions placed the fixation mark so that targets were nearer to the fovea than the locus of best acuity. Indeed, on one of these occasions (the second) the observer adjusted the fixation point
152 t.5 1.4 1.3_ of: ~ 1.2 C, 0 1.1 1 a0 0.9 0.8 1 ~ ~ 1: 0 2 4 6 8 10 o 2 4 6 8 10 ECCENTRICITY (DEG) FIGURE 3 Extrafoveal visual acuity thresholds (MAR = minimum angle of resolutions for illuminated Landolt rings of 0.004 cd/m2 are plotted on a logarithmic axis for two observers. Targets were presented in the right visual field (to the right eye of the observer whose data are on the left and to the left eye of the observer whose data are on the right). Arrows mark the eccentricities that each observer selected to achieve the best discriminability of acuity targets. so that the acuity targets fell within the relatively scotomatous area bordering the fovea. What is worth highlighting from these data is that these two obser- vers (and even the same observer on different occasions) selected dif- ferent retinal loci at which to image scotopic acuity targets, despite the fact that both observers had the best acuity at about the same retinal locus--5 or 6 degrees eccentrically. It would be useful to know on what basis or bases these angles of eccentric regard were selected. Because the acuity gradient is quite shallow at this luminance level, it must be presumed that other information or biases are being used, or at least contribute to the determination of the eccentric viewing locus. One possibility is that the peripheral locus used to view scotopic tar- gets represents an averaging of several tendencies, each of which may drive the eye toward a different retinal locus. For example, the use of perceived brightness of the targets as a criterion might contribute to shifting the viewing locus further peripherally, toward a site where brightness might be supposed to be greatest. On the other hand, the observer might attempt to image targets as near to the fovea as possible
while maintaining their visibility. This latter bias is one which appears to influence many patients who lose central vision in both eyes, even when visual functioning might be served better by the use of a more peripheral retinal locus (Weiter et al., 1984~. Evidence that normal observers may also behave this way when instructed to fixate a scotopic target comes from eye movement recordings which show that the targetts image tends to drift toward and sometimes into the scotomatous foveal region (Steinman and Cunitz, 1968~. The existence of biases, or the use of visual criteria unrelated to the task in selecting an eccen- tric point of regard, are potentially very powerful, since neither patients who lose central vision (White and Bedell, 1985) nor normal observers required to image targets on the optic nerve head (Yap et al., 1986) learn readily to aim the eye to one side of a target by a fixed angular amount. INFLUENCE OF OPTICAL BLUR ON ACUITY AT LOW LUMINANCES The effect of optical blur is to reduce contrast, snore so for fine than for coarse detail. Since at low luminances only coarse detail is resolvable, it might be expected that blur would have little or no effect on acuity. However, review of the literature reveals incomplete and somewhat conflicting information. In 1932, Ferree and Rand measured the effect on acuity of small amounts of astigmatic blur (up to only 0.25 diapers [D]) at illuminations ranging from 0.015 to 36 foot-candles (providing luminances equal to about 0.04 to 100 cd/m2 on their test surface). Their results show that small amounts of blur adversely in- fluence visual acuity at all luminances examined and, surprisingly, that acuity is influenced by the greatest amount at the lowest illumination. Sloan (1968) published data on the effect of blur (produced by crossed cylinders) on the acuity for letters at a range of photopic luminances (approximately 0.5 to 320 cd/m2) and concluded that the deleterious effect of blur diminishes as luminance is reduced. However, if Sloan's data are replotted on a logarithmic axis, the influence of blur is found to remain almost constant across luminances. Takahashi's (1965) data also reveal comparably adverse effects of blur on acuity at a moderate (ca. 320 cd/m2) and a low (3.8 cd/m2) photopic luminance, despite the fact that a small artificial pupil was used. Recently, the effect on acuity of +1.25 D of spherical blur was examined at luminances ranging from photopic to low mesopic levels (Simpson et al., in press). Acuity was measured monocularly for 4-position Landolt rings embedded in a background of proportionately spaced tumbling E's (From, 19661. The results show that this moderate amount of blur reduces acuity at all luminance s, although somewhat less at the lower than the higher luminances (Figure 4~. It is noteworthy that the addition of a +1.25-D lens degrades acuity substantially at the lowest luminance (0.017 cd/m2), even though the unblurred acuity at this luminance is worse than acuities with the lens in place at higher luminances. Of the four observers (all in their twenties), whose data are shown in Figure 4, two (KA and TS) were essentially emmetropic and two were
154 .3 1.t 0.9 0.7 as ~ 0.5 o 0.3 . KA o.e :;; ~`3 0.5 0.3 0.1 - \ 0. -o., Am -o. -~.3 1.3 ~ 1. 1 _ 0.3 r 0. 1 0.9 0.7 ce c, o.s o , . . . . 0.017 0.17 1.7 17.0 170 LUMINANCE (cd/m 2 ~ \ Q \ ~TS l -0.3 1.3( 1.1 0.7 c: ~ o.s o it_ - ~0 . . 0.017 0.17 1.7 17.0 170 LUMINANCE (cd/m 2 ) o.s \ 0.3 0.1 - O. 1 _ -O. 1 -0.3 -0.3 DC \ - Q ~` , . . . . 0.017 0.17 1.7 17.0 170 0.017 0.17 1.7 17.0 170 LUMINANCE (cd/m2 ) LUMINANCE (cd/m2 ) FIGURE 4 Visual acuity (}4AR = minimum angle of resolution) is plotted on a logarithmic axis as a function of target luminance for four obser- vers. Open symbols indicate acuities with best refractive correction in place; solid symbols are the acuities with +1. 25 diopters of added spher ical blu r .
155 myopic (DC, -3 .25 -1.25 x005; TV, -0. 50 -0. 50 x030) . From another study ( in conjunction with Karen Wolf-Kelly of the New England College of Optometry), these four subjects' resting levels of accomodation in darkness (Leibowitz and Owens, 1975) are known: one emmetrope and one myope had relatively near dark foci (KA, 1.7 D; TV, 1.6 D) and the other emmet rope and myope had relatively far dark foci (TS, 0.4 D; DC, O. 6 D) . At low luminances, a shift of accommodation toward its resting level would be expected to affect acuity as does a blurring lens (Johnson, 19761; furthermore, the effect of accommodative error and a blurring lens should summate. However, for these four observers, the resting level of accommodation in darkness was not related strongly to either the unblurred acuity at low luminances or the acuity reduction produced by the blurring lens (Figure 4~. Perhaps the rate at which accommodation shifts toward the dark focus as luminance is reduced varies from individual to individual. A possible explanation for the substantial reductions of acuity by blurring lenses at low luminances is that papillary dilation at dimmer light levels permits such a lens to increasingly impair targets' optical quality. This possibility was addressed in con junction with Trefford Simpson, Ralph Barbeito, and David Loshin of the University of Houston. Contrast sensitivity functions were measured at photopic ( 120 cd/m2) and low mesopic (0.01 cd/m2) space-averaged luminances with and with- out +1.25 D of optical blur. Observers were dilated and cyclopleged, with best correction provided for the 57-cm viewing distance. Viewing was through a 6-mm-diameter artificial pupil, and sensed on a large (22 by 30 degree) Joyce monitor. Ve were presented for 750 ms with a slow onset and offset corresponding to 79 percent correct were determined by averaging six staircase reversals (0.5-dB step size) obtained with a two-alternative, temporal forced-choice paradigm. The results for one observer, shown in Figure 5, demonstrate that the blurring lens reduced contrast sensitivity for mediums and high- spatial-frequency gratings at both low and high luminances. Hence, these data are consistent with the acuity results {Simpson et al., in press) in showing that the blurring lens produces roughly similar shifts of the resolution limit (the extrapolated high-spatial-frequency cutoff) at both luminances. If the effect of blur on contrast sensitivity is compared at spatial frequencies between 0.25 and 2 cycles/degree, it is seen that greater losses of contrast sensitivity occur when the lumi- nance is low. Furthermore, the blurring lens has its first deleterious effect on contrast sensitivity for a coarser spatial frequency at low than at high luminance (about 0.75 cycles/degree at the lower luminance versus about 1.5 cycles/degree at the higher luminance). Since these data were collected with accommodation paralyzed and an artificial pupil in place, the optical properties of the eye-blurring lens combination must be identical at the two luminance levels. It is important to realize that through a 6-mm pupil a +1.25-D lens produces a nontrivial reduction of optical modulation, even for gratings on the order of 1 cycle/degree (Charman, 19833. Thus, one must ask why pho- topiccontrast sensitivity is not affected by the blurring lens at 0.75 and 1 cycle/degree rather than why mesopic contrast sensitivity is gratings were pre- rtical gratings . and thr esholds
156 100 30 - - - cn z go 10 In fir z o 3 1 /4' ~ - - ~ ~ ~ A` A/ _ ~ i/ \ \\ b ~\\\ ~ - \ 0.25 0.50 1.0 2.0 4.0 8.0 16.0 SPATIAL FREQUENCY (c/deg) FIGURE 5 Contrast sensitivity functions shown for one observer at a photopic (open symbols) and a low mesopic (solid symbols) luminance. At each luminance, contrast thresholds were determined with best refractive correction in place (squares) and with +~.25 D of added spherical blur (circles). Viewing was monocular through a 6-mm artificial pupil. affected at the same spatial frequencies. The answer is not yet known, but there is at least one reasonable guess: even though pupil size was physically the same at both luminances, the much greater magnitude of the photopic than the scotopic Stiles-Crawford effect (van Loo and Enoch, 1975) could have produced a significantly smaller effective pupil size for the high-luminance gratings. However, in addition it
157 is suspected that neural processing at photopic luminances, which pro- vides more powerful lateral inhibitory interactions than at low mesopic or scotopic luminances (Robson, 1980), may render the visual system less sensitive to the effect of blur on spatial frequencies below the peak of the contrast sensitivity function. At present a mathematical model is being constructed that includes both optical and neural factors in order to evaluate their relative contributions to the contrast sensitivity losses produced by blur at high and low luminances. RECOMMENDATIONS Although previous investigations have determined profiles that char- acterize how acuity changes at different retinal locations as luminance is decreased, comparably systematic investigations have not yet been carried out for many other functions, such as precision of visual direc- tionalization. Determination of such profiles should be undertaken for a variety of visual tasks and for luminances ranging from photopic to moderate or low scotopic levels in a single group of normal observers. The results will specify the retinal loci at which best sensitivity is attained for each measured visual function and luminance level, as well as the extent of interobserver variability. In addition, the retinal locus that each of the same observers uses to regard targets when fixa- tion is unconstrained should be measured with an unobtrusive, objective technique for sensing eye position. Together, retinal sensitivity pro- fi~e~ and eye fixation measures should help in evaluating the factors that determine which eccentric retinal loci observers adopt to regard scotopic targets. Such information is necessary to develop viewing strategies for the optimal performance of particular visual tasks at dim light levels. The influence of optical degradation produced by papillary dilation and inappropriate accommodation on visual performance at dint luminance levels needs to be investigated more fully. Small refractive errors, such as the astigmatism that remains uncorrected by soft contact lenses, are likely to impair visual acuity and contrast sensitivity substan- tially at low luminance s. Because aberrations of the retinal image ae- pend on the image's wavelength composition, spectral characeristics of the visual tarets or display will have to be considered. Vergence pos- ture as well as accommodation should be objectively and continuously monitored to evaluate how interactions between vergence and accommo- dation contribute to focusing errors at dim levels of illumination. These investigations will reveal the contribution of optical factors to degraded visual performance at low luminances and the extent of individ- ual variability; the results will permit an assessment of when and for whom optical correction would significantly improve visual functioning.
158 REFERENCES Aulhorn, E. 1964 Uber die Beziehung zwischen Lichtsinn und Sehschaerfe. Albrecht van Graefes Archiv fuer Ophthalmologie 167:4-74. Baker, K.E. 1949 Some variables influencing Vernier acuity. Journal of the Optical Society of America 39:567-575. Beck, J., and T. Halloran 1985 Effects of spatial separation and retinal eccentricity on two-dot Vernier acuity. Vision Research 25:1105-1111. Bedell, H.E., M.H. Johnson, and R. Barbeito 1985 Precision and accuracy of oculocentric direction for targets of different luminances. Perception and Psychophysics. Berry, R.N., L.A. Riggs, and C.P. Duncan 1950 The relation of Vernier and depth discriminations to field brightness. Journal of Experimental Psychology 40:349-354. Charman, W.N. 1984 The retinal image in the human eye. Pp. 1-50 in N.N. Osborne and G.J. Chader, eds., Progress in Retinal Research, Volume 2. New York: Pergamon Press. Enoch, J.~., and R.A. Williams 1983 Development of clinical tests of vision: Initial data on two hyperacuity paradigms. Perception and Psychophysics 33:314-322. Ferree, C.E., and G. Rand 1932 The ef feet of intensity of illumination on the acuity of the normal eye and eyes slightly defective as to refraction. American Journal of Psychology 34:244-249. Flom, M.C. 1966 New concepts on visual acuity. Optometric Weekly 57~28~:63-68. Homer, D.G., A.D. Paul, B. Katz, and H.E. Bedell 1985 Variations in the slope of the psychometric acuity function with ecu ity threshold and scale. American Journal of Optometry and PhYsioloqical OPtics 62: 895-900. Johnson 197 6 E f f ects of luminance and st imulu s d i stance on accommodat ion and visual resolution. Journal of the Optical Society of America 66 :138-142. Johnson, COA. r and H.W. Leibowitz 1979 Practice ef fects for visual resolution in the periphery. Perception and Psychophysics 25:439-442. Kerr, J.T. 1971 Visual resolution in the periphery. Perception and Psycho- physics 9:375-378. Kinney, J.S. 1968 Clinical measurement of night vision. Pp. 139-152 in M.A. Whitcomb and W. Benson, eds., The Measurement of Visual Function. Washington D.C.: National Research Council, National Academy of Sciences.
159 Klein, S.A, and D.M. Levi 1985 Hyperacuity thresholds of 1 see: Quantitative predictions and empirical validation. Journal of the Optical Society of America, Pa rt A 2:1170-1190. Koenderink, J.J., M.A. Bouman, A.E. Bueno de Mesquita, and S. Slappendel 1978 Perimetry of contrast detection thresholds of moving spatial sine wave patterns. IV. The influence of the mean retinal illuminance. Journal of the Optical Society of America 68:860-865. Leibowitz, H.W. 1955 Some factors influencing the variability of Vernier adjustments. American Journal of Psychology 68:266-273. Leibowitz, H.W., N.A. Myers, and D.A. Grant 1955a Radial localization of a single stimulus as a function of luminance and duration of exposure. Journal of the Optical Society of America 45:76-78. 1955b Frequency of seeing and radial localization of single and multiple visual stimuli. Journal of Experimental Psychology 50:369-373. Leibowitz, H., and D.A. Owens 1975 Night myopia and the intermediate dark focus of accommodation. Journal of the Optical Society of America 65:138-142. Levi, D.M., S.A. Klein, and A.P. Aitsebaomo 1985 Vernier acuity, crowding and cortical magnif ication. Vision Research 25:963-977. Low, F.N. 1951 Peripheral visual acuity. Archives of Ophthalmology 45: 80-99. Mandelbaum, J., and L.L. Sloan 1947 Peripheral visual acuity. With special reference to scotopic illumination. American Journal of Ophthalmology 30:581-588. Ogle, K.N. 1932 An analytical treatment of the longitudinal horopter; its measurement and application to related phenomena, especially to the relative size and shape of the ocular images. Journal of the Optical Society of America 22: 665-728. Pirenne, M.H ., F.H .C. Marriott, and E.F . O'Doherty 1957 Individual differences in night-vision eff iciency. Medical Research Council Special Report Ser ies, No. 294. London: Her Maj esty ' s Stat ione ry Of f ice . Prince, J.H., and G.A. Fry 1956 The effect of errors of ref. raction on visual acuity. Amer~can Journal of Optometry and Archives of Amer ican Acadernly Optometry 33 :353-373. Rattle, J . D . 1969 Effect of target size on monocular f ixation. Optica Acta 16: 183-192 . Robson, J.G. 1980 Neural images: The physiolog ical basis of spatial vision. Pp. 177-2~4 in C.S. Harris, ea., Hillsdale, li.J.: Lawrence Er lbaum.
160 Sansbury, R.V., A.A. Skavenski, G.M. Haddad, and R.M. Steinman 1973 Normal fixation of eccentric targets. Journal of the Optical Society of America 63:612-614. Schlaer, S. 1937 The relation between visual acuity and illumination. Journal of General Physiology 21:165-188. Sampson, T.L., R. Barbeito, and H.E. Bedell 1981 The effect of optical blur on visual acuity for targets of different luminances. Opthalmic and Physiological Optics 6:279-281. Sloan, L.L. 1950 The threshold gradients of the rods and the cones: In the dark-adapted eye and in the partially light-adapted eye. American Journal of Ophthalmology 33:1077-1089. 1968 The photopic acuity-luminance function with special reference to parafoveal vision. Vision Research 8:901-911. Steinman, R.M., and R.J. Cunitz 1968 Fixation of targets near the absolute foveal threshold. Vision Research 8:277-286. Sullivan, G.D., K. Oatley, and N.S. Sutherland 1972 Vernier acuity as affected by target length and separation. Perception and Psychophysics 12: 438-444. Takahashi, E. 1965 The use and interpretation of the pinhole test. Optometric Weekly 56~18~:83-86. Ten Doe ssc hate, J. 1949 Extra-foveal scotopic absolute threshold and the distribution of ret inal rods. OE>hthalmolog ice 117:110 -115. van Loo, J.A ., and J.M. Enoch 1975 The scotopic Stiles-Crawford effect. Vision Research 15 :1005-1009. von Helmholtz, H. 1925a Treatise on Physiological Optics, Volume 2. J.P.C. Southall, ea., p. 369. Menasha, WI: Optical Society of America. 1925b Treatise on Physiological Optics, Volume 3. J.P.C. Southall, ea., p. 168. Menasha, WI: Optical Society of America. Weiter, J.J., G.L. Wing, C.L. Trempe, and M.A. Mainster 1984 \7isual acuity related to retinal distance from the fovea in macular disease. Annals of Ophthalmology 16:174-176. Westheimer, G. 1979 The scaling of visual acuity measurements. Archives of Ophthalmology 97: 327-329. 1982 The spatial grain of the perifoveal visual field. Vision Research 22 :157-162. Westheimer, G., and S.P. blcKee 1977 Integration regions for visual hyperacuity. Vision Research 17:89-93. White, J.M., and H.E. Bedell 198 5 Eye movements in patients with macular disease. Investigative Ophthalmology and Visual Science 26(Suppl.~:259 (Abstract).
161 Yap, Y .L., H .E . Bedell, and P .L . Abplanalp 1986 Blindspot " fixation" in normal eyes: Implications for eccentr ic viewing in bilate ral macular disease . Amer ican Journal of Optometry and Physiological Optics 62 :259-264. Yap, Y.L., D.L. Levi, and S.A. Klein 1985 Bisection thresholds in central and per ipheral vision. Investigative Ophthalmology and Visual Science 26(Suppl.) :138 (Abstract) .
CONTRAST SENSI TIVITY Lewis O. Harvey, Jr. OVE EVIEW Visual acuity is based on the size of the smallest detail in a visual target (optotype) that permits some criterion level of identi- fication or detection performance (75 percent correct, for example). The smaller the size of this critical detail, the better the vision of the observer. The value of visual acuity measurements is well proven for correcting refractive errors. Yet, under some conditions, individ- ual variation in standard visual acuity measurements often is not able to predict individual variation in performance on some visual tasks, such as target detection and identification (e.g., Ginsburg et al., 1983). A considerable body of empirical knowledge has been gained about the stimulus factors, such as size, exposure duration, contrast, and adaptation level, which influence detection of simple disk-shaped tar- gets (Graham and Margaria, 1935; Lukiesch and Moss, 1940; and Blackwell, 1946~. Although these data, under some circumstances, do quite well in predicting visibility of more complex targets, they often are inadequate in predicting recognition and identification of these targets. In addi- tion, individual differences in performance with these simple targets are not easily related to any measured characteristics of vision, nor are they related by any satisfactory theoretical framework. What is needed is a universal language of vision. Such a language would pro- vide a description of both visual stimuli and the characteristics of the visual system. It would allow us to understand the relationship between basic visual abilities on the one hand and performance with more complex visual stimuli (detection, discrimination, recognition, and identification) on the other. Finally, it would allow under- standing of individual differences in perception of complex visual stimuli in terms of individual differences in basic visual capability expressed in this universal language. These comments were abstracted from the report of Working Group 57 (Committee on Vision, 1985~. 162
163 In the past two decades, a new method of assessing vision has emerged which may provide this universal language. This method is the measurement of the contrast sensitivity function (CSF), and for some purposes, described below, it complements visual acuity. The CSF is typically measured using sinusoidal grating patterns as targets. This use of sine wave gratings was first introduced in vision by Schade (1956) and were subsequently used by early investigators to measure basic visual sensitivity (Westheimer, 1960; DePalma and Lawry, 1962; Campbell and Robson, 1968~. Sinusoidal gratings vary in frequency, contrast, and phase. The number of light-dark cycles of the grating which subtend a 1 degree visual angle is a measure of the spatial frequency of the grating, expressed in cycles per degree (cpd). The human visual system is able to detect spatial frequencies up to about 60 cpd. These is no lower limit but generally measurements are not made below about 0.1 cpd, often because of practical limits of display size. Borrowing the term octave (a doubling of frequency) from audition, the range of spatial frequencies usually measured by the contrast sensitivity function is about 10 octaves. A low spatial frequency consists of broad black and white bands; a high spatial frequency grating has thin black and white bands. Spatial frequency is therefore related to the size of conven- tional objects. When viewing distance and slant are held constant, higher spatial frequencies correspond to smaller objects. The contrast of sinusoidal grating is based on the maximum luminance (LmaX) and the minimum luminance (Lmin) in the grating. It is a dimen- sionless variable with values ranging from 0.0 (a uniform field) to 1.0, the maximum possible. The phase of the grating measures its position in space relative to some predetermined reference point. The minimum contrast at which a grating can be distinguished from a uniform field with some fixed level of accuracy is the contrast thresh- old. The reciprocal of the contrast threshold is called contrast sen- sitivity. The contrast sensitivity function is obtained by measuring contrast thresholds over a range of spatial frequencies. A typical pho- topic contrast sensitivity function is shown in Figure 1. The important features of the contrast sensitivity function seen in Figure 1 are that there is a range of spatial frequencies around 2 to 5 cpd where sensi- tivity is maximum. Sensitivity falls off for lower spatial frequencies and falls off rapidly for higher spatial frequencies. Eventually a high spatial frequency is reached which requires a contrast of 1.0 to detect (the high-frequency cutoff). Spatial frequencies higher than this cutoff frequency cannot be detected by an observer. Because it is measured in terms of the smallest identifiable, high- contrast target and because small sizes correspond to high spatial fre- quencies, visual acuity measures visual sensitivity largely in the higher frequency regions of the contrast sensitivity function. In brief, visual acuity is measured in terms of the size of the critical detail (stroke width of the Snellen letter, for example), but this feature is not the only important one. Snellen acuity letters corres- ponding to acuity of 1.0 have a height of 5 min of arc. The spatial frequencies necessary (but not sufficient) for correct identification of these small letters fall in the approximate range of from 18 to 30
164 , _- 50 ._ ._ - ._ c a 10 _ 0 _ 5 r2 _ _ 04 . ~I I I I I I ~ 1111 ~t ~ 1151 ~t_ 1~ O OS 0 1 0.5 14 5 10 50 SPOtIQI freqve~ (c/deg) FIGURE 1 Photopic contrast sensitivity function of the human visual system for sinusoidal gratings. Both coordinates are logarithmic. Source: Campbell and Robson (1968~. cpd (Ginsburg, 1981a). This is the range of critical spatial frequen- cies necessary for identification of letters at a visual acuity of 1.0. Does the measurement of sensitivity within this range of spatial fre- quences (as with visual acuity) adequately describe the rest of the con- trast sensitivity function? Extensive psychophysical data, including individual differences in contrast sensitivity functions, independence of contrast thresholds at different spatial frequencies, and masking and adaptation experiments, lead to the conclusion that the answer is no. Figure 2 illustrates three different contrast sensitivity func- tions from three Air Force pilots with visual acuities of 1.33, 1.00, and 0.80. The visual acuities of the pilots are predicted from the contrast sensitivity in the high spatial frequency range. The higher the high-frequency sensitivity, the higher the visual acuity. Note the wide variations in low- and high-frequency sensitivity and that a low sensitivity at high frequencies does not necessarily imply a low sen- sitivity at low spatial frequencies. Visual acuity measurements, which are related primarily to high spatial frequency sensitivity, cannot predict contrast sensitivity to low spatial frequencies because threshold to spatial frequencies sep- arated by more that about a factor of 2 (one octave) are statistically independent of each other (Blakemore and Campbell, 1969; Graham and Nachmias, 1971; Sekuler et al., 1984~. This independence of widely separated spatial frequencies is consistent with a model of the visual system containing separate mechanisms, each of which is selectively sensitive to a limited range of spatial frequencies (Campbell and
165 1000 Coo 400 300 200 100 - z Lo in in c: 50 40 30 20 on lo _ Is 4 3 _ 2 _ 1 . C.D. 20/15 x B. W. 20/25 M. R. 20/20 2/4i7g f , , , ~ I I ~ ~ I ~ I I 0.3 0.5 1 2 3 4 5 10 15 20 304051) (30 SPATIaL FREQUENCY (CYCLES PER DEGREE) FIGURE 2 The contrast sensitivity function of three pilots with visual ecu ities of 1.33, 1. 00, and 0. 80. Note the variation of sensitivity below 7 cpd. Source: Ginsburg, 1981b. Robson, 1968; Blakemore and Campbell, 1969; Graham and Nachmias, 1971; Stromeyer and Julesz, 1972~. The contrast sensitivity function has the potential-of adding more information about the functioning of the visual system than that given by visual acuity, because it assesses sensitivity over a wide range of spatial frequencies, while visual acuity pr imar fly measures sensitivity at the high spatial frequencies. While visual acuity cannot predict the spatial contrast sensitivity function in people with abnormal vision, visual acuity also cannot pre- dict contrast sensitivity in people with assumed normal vision. There are individual differences in contrast sensitivity, as well as changes in the contrast sensitivity function with age (Ginsburg, 1981a, 1984; Owsley et al., 1983~. There is some experimental evidence that under some conditions peak contrast sensitivity may be more important than visual acuity for predicting detection and identif ication of ob jects (G insbu fig, 19 81b) . Contrast sensi tivity functions are measured with sinusoidal grating patterns. Although contrast sensitivity functions could be measured with a wide assortment of targets differing systematically ire size and
166 contrast, the human visual system seems to be especially sensitive to sinusoidal targets (Gush and McNelis, 1969; Watson et al., 1983; Ginsburg, 1984~. These sinusoidal gratings have important mathematical properties which allow the application of linear systems analysis to the human visual system. This approach allows both the visual stimulus and the visual system to be described with the same language: that of sinusoidal spatial frequencies. Basic Factors in Spatial Contrast Sensitivity Contrast The dependent variable which is the basis of the contrast sensi- tivity function is the contrast of the sinusoidal test grating. The cost commonly used definition of contrast (C) is: L - L i LmaX ~ Lmin where LmaX is the maximum luminance at the peak and Lmin is the minimum luminance at the trough of the sine wave. Contrast may vary from 0.0 (a field of uniform luminance) to 1.0 (where Lmin = 0~0 and LmaX = 2 ~ O * Lmean) e It is not possible to have a real sinusoidal grating of contrast greater than 1.0, because luminance cannot be less than 0.0. At this time there are no standards for the measurement of the con- trast sensitivity function, although there is general understanding of the factors that influence its shape. These factors are discussed below. Mean Luminance The mean luminance of the sinusoidal gratings has a profound ef feet on the CSF. At high photopic levels of mean luminance, the normal CSF has a peak sensitivity at about 5 cpd and a high-frequency cutoff at about 60 cpd . As mean luminance is lowered, not only do the frequencies of the peak sensitivity and the high-frequency cutoff become lower but the height of the peak also is reduced. At the mesopic level of mean luminance the peak in the CSF function has practically disappeared. This peak, which is so prominent at high luminances, is generally believed to reflect the dynamic interaction between excitatory and inhibitory influences in the visual system. Retinal Locus Compared with the CSF measured at the fovea, the CSF measured for eccentric retinal loci shows a progressive shift to lower spatial fre- quencies for both the high-frequency cutoff and the frequency of peak
167 sensitivity. In addition, there is a progressive lowering of overall contrast sensitivity. This change in the CSF is believed to reflect the unequal way in which the retinal visual field is projected onto the visual cortex (Daniel and Whitteridge, 1961; Schwartz, 1980~. When the contrast sensitivity function is measured at different retinal eccen- tricities with sinusoidal gratings, the size and spatial frequency of which have been adjusted to stimulate equal amounts of visual cortex, the CSF is approximately the same at all retinal loci (Virus and Rovamo, 1979; Rovamo and Virus, 19791. Field Size Visual sensitivity at a particular spatial frequency depends on how many cycles of the sine wave grating are included in the pattern. Sen- sitivity increases as more cycles are included, up to about 10 complete cycles. Usually the low-frequency limit of testing is limited by the largest possible f ield. For example, to measure maximum sensitivity at 0.5 cpd, one would need to test a field 20 degrees square to have 10 complete cycles. Temporal Characteristics Both the exposure duration of the grating target and the time course of its onset influence. the CSF, especially at the lower spatial f requencies. Figure 3 shows the CSFs measured with two different stimr ulus duration characteristics. Notice that with brief exposure duration or with rapid onset of the grating stimulus, sensitivity to low spatial frequencies is enhanced compared with that to longer duration, more gradual onset stimuli. If gratings are modulated sinusoidally in time, the temporal frequency influences the measured contrast sensitivity function. The interaction between temporal and spatial frequency are represented by a contrast sensitivity surface (Kelly, 1979~. Notice that contrast sensitivity at low spatial frequencies is higher when the grating is flickered at a relatively high temporal modulation compared with that at a low temporal modulation. Orientation The visual system is more sensitive to horizontal and vertical gratings than to other orientations; this is an example of the oblique effect (Appelle, 1972), in which horizontal and vertical orientations are more important in vision than are oblique orientations. The con- trast sensitivity, especially for the high spatial frequencies, is reduced for oblique orientations relative to that for horizontal and vertical orientations. This orientation anisotropy may be important for calculations of the effective visual stimulu s.
168 t 000 ~180 z llJ ~10 z o 1 9' ""'I ' ' ' '""I ' ~ . , I ~ ~ lIll ~ t ~ ~ tItl .3 1.0 10.0 50.0 SPATIAL FREQUENCY IN CPD FIGURE 3 Contrast sensitivity functions for long exposure time with gradual onset and offset (sustained presentation) and for short expo- sure time with rapid onset and offset (transient presentation). Note that low-frequency sensitivity is enhanced and high-frequency sensi- tivity is reduced with transient presentations. Gabor Functions There are sets of elementary signals other than sinusoidal gratings which can be used to measure the sensitivity of the visual systems for the purpose of linear systems analysis. One such stimuli is a sinusoi- dal grating which has been multiplied by a Gaussian function. These stimuli are called Gabor functions, after Dennis Gabor who, in 1947, proposed that they could be used as a set of elementary signals for linear systems analysis. Gabor showed that there is a trade-off between localizing a stimulus in space and localizing it in frequency. For example, a point in space is perfectly localized in space but come pletely unlocalized in frequency, because it contains only one fre- quency, but it is completely unlocalized in space because it extends to infinity. Gabor proved that the stimuli maximize the joint localiza- tion in both space and frequency simultaneously. Marcelja (1980) first suggested that cells in the visual cortex have receptive field sensitivity profiles that are of the form of Gabor functions, and further electrophysiological measurements of the visual cortex in monkeys support this idea (Kulilowski et al., 1982; Pollen and Ronner, 1982; Pollen et al., 1984~. Psychophysical evidence sug- gests that the human visual system may also contain mechanisms with characteristics of Gabor functions (Daugman, 198Q; MacKay, 1981; Watson et al., 1987; Pollen et al., l984~. These developments may have conse- quences for the way in which the human contrast sensitivity function is measured, but it is too early to know with any certainty what they are.
169 PRACTICAL IMPLICATIONS FOR PERFORMANCE Because the image of any object can be described as a set of spatial frequencies at various orientations, amplitudes, and phases, there is the potential that an observer's contrast sensitivity function can be used to predict visual performance with more complex visual material. The working group of the Committee on Vision (1985) evaluated the evi- dence that the contrast sensitivity function can predict visual perfor- mance with complex stimuli, and found that considerably more research in this area is needed. If the potential of these early experiments is verified by further studies, the working group of the Committee on Vision believes that there will be a powerful way of studying individual differences in vision, and accounting for some of the variability of individual performance on a wide variety of tasks that are primarily dependent on vision. There is some experimental evidence which suggests that the contrast sensitivity function can predict certain types of visual performance better than can other measures. In one series of experiments, subjects were asked to judge the visual similarity between all possible pairs of random complex grating patterns. These studies found that the similar- ity judgments were accurately predicted by using the spatial frequency content of the gratings in conjunction with the human contrast sensi- tivity function (Harvey and Gervais, 1978, 1981~. In another study subjects were required to identify 6-min-arc-high letters of the alphabet presented for 30 ms (Gervais, 1978; Gervais et al., 1984~. Gervais and coworkers tried several models of visual processing to predict the pattern of identification confusions found among all 26 letters of the alphabet. The model with the greatest predictive power was that based on the amplitude and phase information in the spatial frequency content of each letter filtered by the human contrast sensi- tivity function. The contrast sensitivity function of infants has been used to successfully predict the amount of time that infants spend looking at different types of visual stimuli. The best predictor of looking time has been the amount of contour in the stimulus, the contour density (Karmel, 1969~. One study demonstrated that when stimuli were equated for contour density, infants still preferred some stimuli over others. These preferences were predicted by the spatial frequency characteris- tics of the stimuli (Banks and Stevens, 1981~. Another study showed that when the spatial frequency content of the stimulus patterns was combined with the infant's contrast sensitivity function, both the infant's looking preferences and looking times were better predicted than by the contour density measure (Gayl et al., 1983~. Finally, individual differences in contrast sensitivity functions may be the basis of ind ividual differences in performance on complex tasks. Ginsburg ( 1981) studied the contrast sensitivity functions of three observers with visual acuities of 1. 00, O. 66, and 0.40. The two subjects with acuity below 1.00 required optical correction but were tested without their glasses. These observers had their contrast thresholds measured for the detection and identification of letters of the alphabet and aircraft of different angular size. The individual
170 differences seen in the detection and identification of letters and air- craft were predictable from the relevant spatial frequencies of those targets required for detection and identification and the individual contrast sensitivity functions. The observer with the highest contrast sensitivity in the middle spatial frequencies also was best at target detection and identification tasks, even though this observer's visual acuity was not the best of the three. A second study of 11 Air Force pilots (Ginsburg et al., 1982) indicated that contrast sensitivity, not visual acuity, predicted simulated air-to-ground target detection. Visual acuity and contrast sensitivity functions were measured under high and low photopic levels of luminance. The correlation between the acuity measures and detection range was not statistically significant. There was a significant correlation (0.83, p 0.01) between detection range and peak sensitivity of the contrast sensitivity functions mea- sured at low photopic levels. In a third study (Ginsburg et al., 1983), 84 Air Force pilots took part in field trials which required them to detect an approaching airplane from the ground. Ten sets of field trials were run under widely different visibility conditions, with about eight pilots participating in each set. Correlations were computed between the distance of target detection and various measures of visual resolution, including visual acuity and the contrast sensi- tivity at six different spatial frequencies. Although the working group of the Committee on Vision found it difficult to arrive at an unambiguous interpretation of the results of this study, it seems that contrast sensitivity measures were better able to predict target detec- tion distance than was clinical visual acuity. These suggestive find- ings point to the great need for a major study relating visual acuity and contrast sensitivity measures to target detection and recognition performance. IMPLICATIONS FOR SCREENING If the potential power of the contrast sensitivity function to measure spatial resolution in vision is supported by future research and is extended to scotopic vision, a powerful screening method could be developed. Based on current information, a large sinusoidal target, compensated for by the cortical magnification factor (the Kelly stimuli look promising), could be the basis for measuring the scotopic contrast sensitivity function. The large target could be centered on the fovea, but because of its size (30 degrees or so) it would cover the area of the retina, which is inhomogeneous in terms of sensitivity and resolu- tion. The cortical magnification factor, which has been defined in terms of photopic vision, may not be applicable to scotopic functions, and thus, different mapping functions should be used. Further research is necessary to demonstrate the validity of such measurements for pre- dicting task performance, but a positive outcome is likely.
171 REFERENCES Appelle, S. 1972 Perception and discrimination as a function of stimulus orientation: The "oblique effects in man and animals. Psychological Bulletin 78:266-278. Banks, M.S., and B.R. Stephens 1981 A Test of the Linear Systems Approach to Studying Infant Pattern Vision. Poster presented at the meeting of the Society for Research in Child Development, Boston, Mass. 81ackwell, H.R. 1946 Contrast thresholds of the human eye. Journal of the Optical Society of America 36:642-643. Blakemore, C., and F.W. Campbell 1969 On the existence of neurons in the human visual system selectively sensitive to the orientation and size of retinal images. Journal of Physiology 203:237-260. Campbell, F.W., and J.G. Robson 1968 Application of Four ier analysis to the visibility of gratings. Journal of Physiology 197:551-566. Committee on Vision 1985 Emergent Techniques for Assessment of Visual Performance. Commission on Behavioral and Social Sciences and Education, National Research Council. Washington, D.C.: National Academy Press. Daniel, M., and D. Whitteridge 1961 The representation of the visual field on the cerebral cortex in monkeys. Journal of Physiology 159:203-221. Daugman, J.G. 1980 Two-dimensional spectral analysis of cortical receptive field profiles. Vision Research 20:847-856. DePalma, J.J., and E.M. Lowry 1962 Sine-wave respone of the visual system. II. Sine-wave and square wave contrast sensitivity. Journal of the Optical Society of America 52:328-335. Gayl, I.E., J.O. Roberts, and J.S. Werner 1983 Linear systems analysis of infant visual pattern prefer- ences. Journal of Experimental Child Psychology 35:30-45. Gervais, M.J. 1978 Spatial Frequency and Letter Confusions: A Test of Fourier Analysis as a Metric and a Model for Form Perception. Master's thesis, Department of Psychology, University of Color ado . Gervais, M.J., L.O. Harvey, Jr., and J.O. Roberts 1984 Identification confusions among letters of the alphabet. Journal of Experimental Psychology: Human Perception and Performance 10:655-666.
172 Ginsburg, A.P. 1981a Spatial filtering and vision: Implications for normal and abnormal vision. Pp. 70-106 in L. Proenza, J. Enoch, and A. Jampolsky, eds., Clinical Applications of Visual Psycho- physics. Proceedings of a Symposium held in San Francisco, October 1978. Cmabridge, England: Cambridge University Press. 1981b Proposed New Vision Standards for the 1980's and Beyond: Contrast Sensitivity. Air Force Aerospace Medical Research Laboratory, Repot AFARML-TR-80-121. Wright-Patterson Air Force Base, Dayton, Ohio. 1984 Visual form perception based on biological filtering. Pp. 53-72 in L. Spillmann and B.R. Wooton, eds.' Sensory Experience, Adaptation, and Perception. Hillsdale, N.J.: Erlbaum. Ginsburg, A.P., J. Easterly, and D.W. Evans 1983 Contrast sensitivity predicts target detection field performance of pilots. Proceedings of the Human Factors Society (27th Annual Meeting):269-273. Ginsburg, A.P., D. Evans, R. Sekuler, and S. Harp 1982 Contrast sensitivity predicts pilots' performance in aircraft simulators. American Journal of Optometry and Physiological Optics 59:105-109. Graham, C.H., and R. Marg aria 1935 Area and the intensity-time relation in the peripheral retina. American Journal of Physiology 113:299-305. Graham, N., and J. Nachmias 1971 Detection of bating patterns containing two spatial frequen- cies: A comparison of single and multi-channel models. Vision Research 11:251-259. Guth, S.K., and J.F. McNelis 1969 Threshold contrast as a function of complexity. American Journal of Optometry and Archives of American Academy of Optometry 46:98-103. Harvey, L.O., Jr., and M.J. Gervais 1978 Visual texture perception and Fourier analysis. Perception and Psychophysics 24:534-542. 1981 Internal representation of visual texture as the basis for the judgment of similarity. Journal of Experimental Psychology: Human Perception and Performance 7:741-753. Karmel, B.Z. 1969 The effect of age, complexity, and amount of contour on pattern preferences in human infants. Journal c] ExE~oi- mental Child Psychology 7:339-354. Kelly, D.H. 1979 Motion and vision. II. Stabilized spatio-temporal threshold surface. Journal of the Optical Society of America 69:1340-1349. Kulikowski, J.J., S. Parcel ja, and P.O. Bishop 1982 Theory of spatial position and spatial f requency relations in the receptive f ields of simple cells in the visual cortex. B iolog ical Cyber net ics 4 3: 187-19 8 .
173 Luckiesh, M., and F.K. Moss 1940 Function adaptation to near vision. Journal of Experimental Psychology 26: 352-356. MacKay, D.M. 1981 Strife over visual cortical function. Nature 289 :117-118. Ma reel j a, S . 1980 Mathematical description of the responses of simple cortical cells. Journal of the optical Society of America 70:1297-1300. Owsley, C., R. Sekuler, and D. Siemsen 1983 Contrast sensitivity throughout adulthood. Vision Research 23: 689-699. Pollen, D.A., and S.F. Ronner 1982 Spatial computation performed by simple and complex cells in the visual cortex of the cat. Vision Research 22:101-118. Pollen, D.A ., M. Nagler, J . Daugman, R. Kronauer , and P . Cavanagh 1984 Use of Gabor elementary functions to probe receptive field substructure of poster for inferotemporal neurons in the owl monkey. Vision Research 24: 233-241. Ro~ramo, J., and V. Virus 1979 An estimation and application of the human cortical magnifica- tion factor. Experimental Brain Research 37:495-510. Schade, O.H. 1956 Optical and photoelectric analog of the eye. Journal of the Optical Society of America 46:721-739. Schwartz, E.L. 1980 Computational anatomy and functional architecture of striate cortex: A spatial mapping approach to perceptual coding. Vision Research 20:645-669. - Stromeyer, C.F., III, and B. Julesz 1972 Spatial frequency masking in vision: Critical bands and spread of masking. Journal of the Optical Society of America 62:1221-1232. Virus, V., and J. Rovamo 1979 Visual resolution, contrast sensitivity, and the cortical magnification factor. Experimental Brain Research 37:475-494. Watson, A.B., H.B. Barlow, and J.G. Robson 1983 What does the eye see best? Nature 302:419-442. Westheimer, G. 1960 Modulaton thresholds for sinusoidal light distributions on the retina. Journal of Physiology 152:67-74.
THE PERCEPTI ON OF THE LAYOUT OF SPACE DURING FLIGHT Ralph Norman Haber Vision is the single sensory modality critical for flight opera- tions. The other senses contribute little (in a positive way) to the information processing needed to operate aircraft. This truism applies as much during night flight as during the day. While instruments and automatic control systems provide the pilot with substantial aids during night flight, they do not replace the pilot. More importantly, virtually all instruments require visual perception and visual atten- tion, both of which are more difficult during night flight. This paper focuses on the visual tasks concerned with space per- ception and spatial orientation that are necessary to pilot aircraft. I have divided the topic into five parts. In the first, I describe several visual performance task analyses. These are descriptions of the tasks that pilots must performr-tasks for which pilots depend on their vision for the information necessary to carry them out. The categories in which these tasks fall include takeoff and landing, navigation, cross-country travel, target acquisition and attack, and defensive maneuvers. The next section focuses on the subset of these tasks that are required during night flight. Section three examines the ways in which the visual system normally changes in its function when light levels decrease, especially with respect to flying perfor- mance tasks. I then distinguish between those changes in visual func- tioning that affect everyone in much the same way and those that are due to large individual differences that interfere with the normal visual functioning of some people during nighttime. It is the indi- vidual differences that are relevant to the development of screening tests to select pilots. Finally, the implications of these analyses for the training of pilots are discussed, especially with respect to research efforts that are needed to meet these implications. My topic is limited in several important respects. I will confine my analysis exclusively to the visual tasks facing a pilot and not be concerned with other crew stations. Other crew members have different performance tasks, but the dependence of those tasks on vision is still overwhelming, and the principles developed here pertaining to the pilot apply to other crew stations as well. Furthermore, I am primarily con- cerned with the visual tasks involved in spatial perception rather than simply form or shape recognition and identif ication. For a pilot, perceiving space means seeing the layout of the scene beneath him, 174
175 including the contours and orientation of the terrain, the objects attached to or above the terrain, their locations on the terrain, his own orientation and rate and direction of motion above the terrain, and his motion relative to the terrain. Finally, while substantial gains have been made on instrumentation to aid day and night flight, I will consider only conditions of visual flight: those in which the pilot must direct his vision outside of the cockpit in order to fly. PERFORMANCE TASKS REQUIRING VISION DURING FLIGHT Every flight manual contains a list of the principal tasks facing a pilot that require visual information and processing. What is missing from the flight manuals is a description of the visual information needed by the pilot for him to perform each of these tasks. In the sections that follow, I have tried to provide the best descriptions of that visual information and visual processing that are presently avail- able. I have drawn heavily on Miller's manual (1983, and subsequent re- visions), which is used to train fighter pilots for low-altitude flight (a vision-demanding task), and on Haber (1986) as well as reports from the substantial current programs of research that are under way by the Air Force and other laboratories. Even so, these descriptions are not yet very complete, in most cases because the nature of the visual infor- mation that is needed is not known, nor is it known how that informa- tion is processed by the pilot. Before taking up indiv~dup1 performance tasks that depend on vis- ion, consider some general characteristics of the visual perception of scenes. When we gaze at a visual scene {either while flying or standing on the ground), we perceive a layout, an arrangement of objects on a continuous ground surface. The ground itself has a contour--it varies in elevation, and does so in a regular or irregular fashion. The ground also has a surface quality: the texture or substance that makes the difference between the visual appearance of water, grass, wheat, desert sand, rock, mountains, and the like. The layout is of objects on the contoured and textured ground surface. Every object has a location on the ground, a location near or far from other objects. Often, the ob- jects form a pattern of arrangements: they are arranged in a row, or form a square, or are along a street. Perceived layout has dimension- ality: objects are near or far from each other, and I (the perceived am near or far from some of them. Sometimes the dimensions are per- ceived in absolute units, so that we perceive that two objects are 100 feet apart, or that I am 500 yards from a tree. More often, we only perceive relative arrangements: one object is nearer than another, but not by so many feet or miles. The objects that we perceive on the ground surface in a scene are themselves things, usually ones that we can identify or recognize. Some are highly familiar--a VW beetle, or my wife; whereas some are only identifiable--cactus, a road. It is rare that we encounter a scene of objects that are totally unidentifiable as anything related to what we have encountered before.
176 I can nearly always locate myself in the scene being viewed, as being located in a particular place in relation to some or all of the objects in the scene, or to a position on the ground surface. I may perceive the layout of the objects from my own perspective view at the moment, although usually I can perceive the layout of the objects in the scene independently of my own location. The positions of the ob- jects do not change in relation to one another as I move through or around the scene. Their relationships among each other, and to the contour of the ground, remains fixed, even as I move. Most of the time I am riot interested in my distance to any particular object: rather I perceive the layout of all of the objects, irrespective of my location. To perceive my distance to some place in the scene may even require focused attention. As a terrestrial being, I perceive the layout of most scenes while I am in the scene, while I am standing on the same ground as the objects of the scene. Occasionally, I gaze at a scene while elevated above the scene, as from a balcony, so that I am not physically attached to the scene I am viewing. In the latter cases, my viewing is not along the textured ground surface but through empty air. This makes it much harder to locate myself in the scene, and especially harder to perceive my distance to any particular object. Air viewing, as distinct from ground viewing, of a scene probably does not make it more difficult to perceive the layout of the arrangements of the objects in the scene it- self; only my location within the scene. These differences between air and surface viewing are important for flying, because pilots invariably view the terrain below them through air, thereby reducing their percep- tion of being located at a particular place in or over the scene. This makes the perception of elevation above the ground and the range dis- tance to objects on the ground difficult. I have tried to describe in brief terms the characteristics of per- ceived spatial layout. This is a somewhat different description than the traditional view of space perception, which is limited primarily to lists of cues to distance or depth, lists that are made in answer to questions about how far away something appears. What is necessary is to ask questions about the entire scene, about the layout of all of the objects in the scene considered as an arrangement. Recent advances in spatial psychophysics (see Haber, 1985a, 1985b, in press for the theoretical developments and empirical examples) have shown that it is possible to study perceived layout directly and to assess a variety of questions about the way we perceive scenes. These questions include discovering the structure of perceived space, the accuracy with which our perceived layout matches the actual arrangements of objects in the scene, the variables that affect perceived layout, the effects of ele- vated or air viewing of scenes, the independence of perceived structure of visual space from the viewing position of the observer, and many more. The above discussion has been abstract and concerned with any per- ceiver looking at any visual scene. Using this information about gen- eral perceived layout as backs round, several very specif ic tasks that conf rant pilots while they f ly , tasks that requ ire spatial perception, will be considered.
177 Landing Table 1 provides a list of the performance tasks requiring visual information during landings. Assuming that the pilot has found the airport, the initial visual tasks required to land include being able to perceive current altitude above ground level, the orientation of the runway, the slope of the ground around the runway, and the appropriate ground track of the required approach traffic pattern. Once on the final approach, the pilot must then perceive the slope of the ground and the attitude of the plane relative to the ground and the alignment of the flight path to the runway, corrected for wind conditions, and acquire sufficient visual information to be able to achieve and main- tain proper aircraft attitude, glide slope, and speed over the course of the final approach, flare, and touchdown. There are three general kinds of visual tasks for the pilot con- tained in this list: perceiving the slope of the terrain and arrange- ments of the objects on the terrain, especially the runways in relation to surrounding terrain, buildings, and obstacles; perceiving the air- craft's location and attitude in relation to those arrangements; and maintaining visual control of the aircraft's movement through the scene. The first two are concerned with perceived layout, whereas the third requires the combination of perceived layout with reafferant response- produced stimulation that permits precise control of movements through space. The best work on a pilot's perceived layout of space has been done by Miller (1983), in which he begins with a classification of terrains in terms of the kinds of visual information that they provide to a pilot. While Miller's work focuses on terrain and obstacle avoidance when flying at low altitude, it is equally applicable to approaches to landing. Miller classifies terrains as varying from richly informative to impoverished along four dimensions: the texture of the surface, the irregularity of the surface, the presence of objects of known size, and the visibility of fine details. Terrain surface textures produce optic flow patterns (and at very low altitudes a sense of ground rush) that specify changes in altitute above the ground and the rate of convergence with the ground. Some of this optic flow is registered only across cen- tral vision, but it is likely that the most important part is to stimu- late the peripheral retina to provide information to maintain orienta- tion (Liebowitz and Post, 1982~. Furthermore, optic flow arising from movement over a textured terrain defines the contours of the terrains including its undulations and discontinuities. Recent laboratory work has shown that from optic flow alone (called kinetic depth in this con- text) the shape of a surface or solid can be perceived; all that is nec- essary is that we move past the object or it moves past us (Braunstein, 1975; Marr, 1982~. Besides a textured ground surface, Miller (1983) specified terrains in terms of the regularity of their surface contours. An irregular ground surface provides dynamic interposition and optic occlusion which specify ground clearance over the obstacles and the irregularities. The third dimension--the presence of objects of known size on the ter- rain--allows the perceiver to scale the distances among objects and the
178 TABLE 1 Landing Tasks Requiring Vision Performance Task Visual Task Source of Visual Information Before final approach: (a) Determine ground track of traffic pattern Perceive layout of scene (b) Enter pattern at Perceive layout proper place and Perceive AGL AGL (c) Execute pattern Perceive layout; while maintaining Perceive AGL change On final approach: Perceive direction a) Align velocity vector to runway by Maintan glide slope and velocity c) Flare attitude before touchdown Perspective of texture gradi- ents from terrain textures and terrain features; dynamic occlusion patterns; motion perspective of light from reflected terrain features; object recognition and iden- tification; stereopsis (unlikely). See above. Optic flow arising from terrain texture; dynamic occlusion of terrain eleva- tion variation; fine detail detection; familiar size of terrain features. See above. Orientation of optic flow; change in rate of optic flow; ground rush; texture density change; fine detail detection; dynamic occlusion. Differential optic flow across the two retinas arising from terrain texture and features; differential optic flow across each reti- nal pattern. Perceive direc- tion; Perceive direc- tion change Perceive change in AGL Perceive change in AGL; Perceive velocity ? See above. Invariants in optic flow patterns from runway edges, lights, and adjacent scene. See above. See above.
179 range to the objects. Finally, the visibility of fine details of the textured ground or of the objects (e.g., pebbles, leaves, or cactus spines) allows the pilot to make absolute size estimates, even in the absence of knowledge of the absolute size of the objects (e.g., if the pilot can see the veins in the leaves, he must be getting closer to the trees). The reafferant processes have not been examined in detail as they concern layout tasks, but recent work has focused on their importance for approach and landing tasks. For example, Warren and Owen (1982) have shown that pilots can maintain their proper glide path and approach attitude from invariants in the optic flow of visual stimulation that reach their eyes during the final approach. These invariants depend on the dynamic changes in stimulation arising from the visual shape of the runway defined by its borders and lighting, as well as from the textured surfaces of the ground, taxiways, and nearby buildings. This means that the pilot makes control input responses to the aircraft that maintain invariant stimulus configurations of the appearance of the scene as he approaches the runway, and that whenever the aircraft deviates from such a flight path, the pilot makes new inputs so as to reestablish the invariances of stimulation reaching his eyes. As a final example, Regan and Beverly (1979) and Kruk and Regan (1983) have examined the sources and processing of information that pilots use to perceive their direction of flight, especially in rela- tion to terrain features, such as their alignment to the runway on final approach. Regan's work has demonstrated that pilots use bino- cularly received changes in stimulation as it differentially flows across the two retinas to provide precise runway alignment and flight path information at the beginning of and throughout the final approach. Thus, when the pilot's velocity vector is through the numbers at the end of the runway, light reflected from the numbers simply expand sym- metrically on both retinas, whereas all other reflected light, such as from buildings along the runway, both expand and drift off to the sides of the retina, with a differential rate of drift between the two eyes, depending on which side of the velocity vector each object is located. The pattern of differential drift when the pilot is approaching objects defines the velocity vector and all objects on a collision course with the approaching plane. The cells of Table 1 have been filled in according to the visual tasks corresponding to each of the performance tasks and the sources of the visual information needed to carry out those visual tasks. Construction of a table like this one (and the ones to follow) demands more knowledge than is presently available about spatial perception, so there are still many guesses and question marks in the tables (I have not attempted to specify the visual information for perceived absolute velocity on final approach). Cross-Country Flight Table 2 lists the principal performance tasks necessary to get an aircraft from one place to another during cross-country flight
180 TABLE 2 Cross-Country Travel Tasks Requiring Vision Pe rformance Task Visual Task Sou rce of Visual Information Maintain altitude (AGL? Perceive AGL See Table 1 Maintain course Perceive horizon Contrast between sky and heading terrain; discontinuity in edges, features, color. Perceive and Contrast between figure and identify object background; familiar shape. Maintain ground track Perceive layout See Table 1 of space Maintain attitude Maintain spatial orientation Perceive horizon See above Perceive AGL change See Table 1 Perceive horizon See above Avoid obstacles and Perceive layout See Table 1 Perceive AGL See Table 1 Maintain formation Perceive Texture density; fine detail; distance; perspective; motion. Perceive rela- ? Live altitude; Perceive rela- Motion perspective; ? Live velocity - Dock with tanker Perceive See above distance; Perceive rela tive attitude; Perceive rela- See above tive velocity; Instrument cross- Near accommo- Near fixation checks change; Adaptation Changes in ambient light change; levels; Central vision Focused and narrow attention
181 (navigation and specific mission-oriented tasks are considered in sep- arate categories below). Once the aircraft is on the proper course, the pilot then must be able to maintain course and altitude above the ground, maintain aircraft attitude and spatial orientation, avoid high elevations and obstacles, maintain formation or contact with other friendly aircraft, dock with tankers for refueling, and perform the necessary periodic instrument cross-checks. Again, while this list can be found in any flight manual, the characteristics of the visual infor- mation and perceptual processing needed to carry out these tasks are only beginning to be understood. When the planned altitude is less than several thousand feet above ground level, the visual information needed to perceive ground clear- ance, heading, and ground track is the same as that discussed in the previous section on landing (see Miller, 1983~. At an altitude of several miles or more above the ground, terrain characteristics provide less and less of the information needed to perceive aircraft heading, ground track, unplanned deviations in altitude, and aircraft attitude. Such information is typically available visually from perceived changes in the position of the horizon, both in its elevation above or below the velocity vector and in its orientation relative to the wings. Both of these sources of information are also represented in round gauges on the instrument panel, and for some aircraft in the configurations of symbols on the head-up display. If the horizon is not visible, either because of weather or extremely high altitude, then only instruments allow the pilot to maintain aircraft course, altitude, and attitude. Although the pilot can navigate successfully while depending exclu- sively on instruments, in the absence of both visual contact with a textured ground and a clearly marked horizon, he is likely to exper- ience difficulties in maintaining spatial orientation--he cannot per- ceive the attitude of the aircraft. This probably results, at least in part, from the fact that the instrument displays stimulate the pilot's central vision, whereas his orientation-sensitive system requires sti- mulation of his peripheral retina (Liebowitz and Post, 19821. Optic flow produced by ground textures and objects provide this continually; instruments provide this rarely, if at all. In addition to frequent out-the-window cross-checks on heading, ground track, altitude above the ground, attitude, and wingmen (and, of course, mission-related checks for enemy aircraft, missiles, or poten- tial ground targetsy, the pilot must make continual visual instrument cross-checks. these require adjustments in accommodation from far to near vision, in light adjustment (often from. extremely high ambient light levels of full sunlight to the much darker levels below the canopy rails of the cockpit), and in adjustment from full-field viewing to an exclusive re fiance on foveal vis ion under visual search control. The remaining items in Table 2 also occasionally occur as part of cross-country flight. If the flight path is low, relative to the terrain, then the avoidance of hills and mountains, as well as tall ground-based obstacles, in the flight path is cr itical. Much of ter- rain avoidance depends on accurate perception of the terrain features and their locations in relation to the projected flight path, using the visual sources of information already described. However, dynamic
182 occlusion of far terrain features by interposed, protruding, nearer ones is a more direct source of information about intrusions in the flight path. If a far terrain feature is progressively covered up by a nearer one as the pilot approaches it, then the nearer protrusion is on a col- lision course with the flight path. Formation flying places quite different visual demands on the pilot, since he must attend to perceptual invariants in the appearance of his wingman's position and orientation, regardless of the magnitude of the control inputs he must make to his aircraft. It is not known yet how much of those invariants are registered on the peripheral retina, as they might be when the pilot is attending to his own aircraft. Rather similar visual information is needed to accomplish midair refueling, except that the perceptual invariants now concern the rates of closure and joining, as well as relative position invariance. Navigation Many of the visual tasks required for cross-country flight are basi- cally navigational performance tasks. However, in Table 3 the special- ized navigation tasks that depend on visual information have been listed separately. Even when computer instrumentation equipment is available, many navigational tasks still depend on visual information acquired from the terrain and horizon. One of those tasks most dependent on vision is being able to identify landmarks on the terrain about which the pilot has previously been briefed, which requires being able to perceive the layout of the terrain and its objects. Furthermore, the pilot must be able to predict his ground track over the terrain, specifically over those parts of the terrain that his present or planned control settings will take the aircraft. He also must be able to perceive the distance range (or the time that will elapse) to reach prominent terrain fea- tures. While Miller (1983) has discussed the visual control of these tasks, relatively little is known about how they are done. Target Acquisition and Attack Table 4 lists the principal mission tasks requiring visual infor- mation that are related to finding and identifying targets, and then the visual control of the maneuvers required to attack them. These include a variety of visual search and recognition tasks, such as perceiving the presence and location of other aircraft, and eventually their identification, and discr imitating potential ground targets from background and camouflage and eventually identifying their properties. After successful visual search performance, visual information is required for the control of the maneuvering of the aircraft in close proximity to the ground (or airborne targets). These tasks include being able to perceive the range to a target and the rate of closure wi th the target. In addition to range and closure rate, the pilot must be able to perceive and predict his g round track, especially his loca- tion fo, lowing a bomb burst. Finally, he must perceive his rapidly
183 TABLE 3 Navigation Tasks Requiring Vision Performance Task Visual Information Source of Visual Information Predict and maintain Perceive layout of space See Table 1 ground track Identify landmarks Perceive and identify See Table 2 objects; Perceive layout See Table 1 Maintain course Perceive horizon See Table 2 heading Perceive and identify See Table 2 objects Predict range or Perceive distance; See Table 2 elapsed time Estimate time ? changing position with respect to the ground in order to control maneu- vers such as when to pull up during for dive bombing or how high he must be to initiate a split-S maneuver. Defensive Maneuvers Table 5 lists some of the principal tasks that require vision for evading detection or, if detected, for evading attack and destruction. In the low-altitude environment, these include a variety of maneuvers involved in terrain masking, in which the irregular features of the terrain are used to mask the visual, thermal, or electronic Signatures of the aircraft. In all environments, these include evasive tactics to break contact with a pursuing missile or attacker. All of these tasks depend on an accurate perception of the layout of the terrain and an accurate perception of the pilot's location with respect to those ter- rain features. Summary Tables 1 through 5 summarize what is known about the visual require- ments of flight. The entries in the left columns in each table are traditional labels found in flight manuals that list the flying tasks to be performed. The center columns note the visual tasks required to carry out each of the flying tasks. The righthand columns attempt to spell out the kinds of visual information potentially available to a pilot for each of those visual tasks.
184 TABLE 4 Target Acquisition and Attack Tasks Requiring Vision Source of Performance Task Visual Information Visual Information Aircraft Perceive and identify identification objects See Table 2 Ground target Perceive and identify See Table 2 identification objects; Perceive layout of space See Table 1 Range and closure rate Perceive distance See Table 2 Perceive change in range ? Perceive layout of space See Table 1 Ground track prediction Perceive layout of space See Table 1 Intercept prediction ? ? AGL for dive pull-up Perceive change in AGL See Table 1 Perceive absolute AGL ? AGL prediction of split S Perceive change in AGL Perceive absolute AGL See Table 1 ? Several kinds of visual information have been distinguished. One concerns identification of fine detail, requiring high visual acuity of the kind assessed by a Snellen eye chart. Another concerns detection of targets against backgrounds, which usually requires good contrast sensitivity, especially in the middle- and low-spatial-frequency ranges (Ginsberg, 1981~. A number of tasks require sensitivity to gradients of optic flow across the entire retina of each eye, and often these flow gradients must be compared between the two eyes. This flow arises from the pilot's motion across terrain with variable patterns of reflectance. Finally, many tasks require that object definition, surface contours, and object location be def ined by shape-from-motion information arising from the pilot' s motion. PERFORMANCE TASKS REQUIRING VISION DURING NIGHT FLYING me above lists and descriptions were made without respect to illu- mination level or atmospheric conditions. Military demands require that operations occur at night and in bad weather, as well as under ideal conditions, and pilots must perform their tasks even under very low light levels. In general aviation as well, personal missions and economics require that many operations be carried out at night.
185 TABLE 5 Defensive Maneuver Tasks Requiring Vision Performance Task Visual Task Source of Visual Information Terrain masking Perceive layout of space See Table 1 Obstacle avoidance Perceive layout of space See Table 1 Jinking Perceive layout of space See Table 1 Ridge crossing Perceive layout of space See Table 1 Afterburner ignition An agressor in pursuit A sense of imminent danger While some flying tasks are altered to fit the realities of night- time operations, nearly all of these tasks must be performed in the same way, to the same close tolerances and high standards, as they are during the day. Perhaps the only tasks that are restricted from the repertoire of an Air Force pilot after dark are low-altitude terrain masking and low-altitude pop-up dive bombing. Therefore, the entire discussion in the preceding section about the visual information that is needed to fly during the day applies equally to the needs of night flying. These similar demands at night create the justification for this paper. CHANGES FROM DAY TO NIGHT IN THE PROCESSING OF VISUAL INFORMATION The amount of available light affects virtually every function of the human visual system. In most cases functioning is so utterly transformed that the principles used to describe high light {photopic) level activities differ radically from those under low light ~ scotopic) levels: the functional neural organization of the visual system is different. Under high light levels, both rods and cones are active, and the neural interconnections display both inhibition and excitation. Under low light levels, the cones cease to function at all, and neural connections are primarily excitatory in nature. These effects funda- mentally change the organization of the receptive processes and the connections within the retina, between the retina and the cortex, and within the cortex. In this section, I briefly describe the perceptual implications of the changes in functioning of the human visual system when ambient light drops from photopic to scotopic levels, especially changes that have an impact on perception of the layout of space and on the visual control of movement through space. Table 6 provides an overall summary.
186 TABLE 6 Examples of Changes in Visual Processing from Day to Night ~ 1) Loss of visual acuity--high spatial f requency sensitivity (2) Reduction in contrast sensitivity--for all spatial frequencies (3 ~ Loss of color (wavelength) discr imination (4 ~ Loss of optic flow f rom terrain texture (5) Loss of texture gradient perspective 6 ~ Ga in of foveal blind spot (7) Altered target search strategies (8 ~ Lowered sensitivity and slowed recovery of sensitivity to light after exposure to high light levels 9 ~ I nc reased n ight myop ia 10 ~ Loss of hor izon 11) Additional sensitivity loss f rom glare 12 ~ Additional sensitivity loss and recovery delay f rom hypoxia 13) Additional sensitivity loss and recovery delay from acceleration forces 14 ~ Increased sky backg round complexity f rom stars 15) Switch from su rf ace textur e to point sou rce illuminations from ground The most dramatic change concerns the ability to resolve fine detail. High-spatial-frequency visual acuity is monotonically depen- dent on luminance, with a lO-fold loss in acuity from the luminance available at high noon to room-level lighting. Below twilight levels, fine-detail acuity is barely measurable (Hecht, 1928~. This kind of acuity, which is needed to resolve the textures of surfaces as well as for object recognition, is mediated almost exclusively by the cones and by the receptive field organization of the retina, which is dependent on a mixture of inhibitory and excitatory interconnections. Similarly, contrast sensitivity to lower spatial frequencies, which is needed to perceive surface textures as well as the presence of a target against a lighter or darker background, also deteriorates sharply as luminance levels decrease. Some natural features in the environment contrast only in hue and not in brightness, but since only the cones are able to dis- criminate among wavelengths of light on the basis of wavelength alone, color discrimination also disappears at night. Taken together, the loss of sensitivity to details, color, and contrast as light levels drop means that surface texture quality and objects on the terrain are no longer perceivable at low light levels. There are also dramatic losses in information that arises from the pilot's motion over the terrain. When the terrain is sufficiently tex- tured, or has sufficient numbers of objects scattered about, variations in the patterns of reflectance produce variations in the distribution of luminances on the retinal images, which produce an optic flow across the retinas as the pilot moves over the terrain. However, at scotopic light levels, these differences in laminates are not discriminable, and there is no optic flow arising from terrain or object textures.
187 This loss is probably the most serious one with respect to the pilot's acquisition of information about the layout of the terrain and his motion over it. One positive change with the arrival of night is the appearance of ground lights. However, even at low altitudes, because of the loss of all detail, color, and contrast, lighted objects on the ground become point sources. This means that while they can differ in their bright- ness, they individually convey no information about their distance. But point sources can be highly informative to a moving observer. A randomly arranged array of point sources attached to a contoured ground surface conveys information regarding the shape of the contours to a moving viewer. This kinetic depth effect (now referred to as shape and depth from motion) was first studied by Wallach and McConnell (1953) and is now well documented by Braunstein (1975) and Ullman (1979), among others. Computer-generated imagery for flight simulators is based on this principle. Depth from motion can theoretically provide a pilot with informa- tion about the contours of the terrain, obstacles and hills above the terrain, and changes in his relative distance from the terrain. How- ever, not much is known about the resolving power of this information: what density of lights at what viewing distance is required to perceive what variation among ground elevations? Of nearly equal seriousness as the loss in surface features is the fact that the pilot loses the horizon at night. This occurs because the contrast difference between sky and land decreases, and the remaining contrast is usually.<below the reduced sensitivity of the pilot's vision at night. This means that a visually marked horizon is effectively invisible to the pilot at night. There are two serious consequences of the loss of the horizon: spatial disorientation and false horizon illusions. Spatial disorientation occurs when a pilot loses positional refer- ences so that he cannot perceive his attitude or changes in attitude. The visual horizon is the main source of information for a pilot about his attitude: it defines a wing-level and nose-level flight attitude and signals immediate deviation from, those attitudes. While a pilot has a small round gauge with an attitude indication (and may even have symbols on his head-up display), instruments are far less effective in signalling attitude than is visual reference to the horizon. In the absence of the horizon, even with functioning instruments, pilots fre- quently report not knowing their attitude or inappropriately relying on vestibular information that cannot be trusted in the absence of vision. Absence of a horizon also leaves the pilot open to a variety of vis- ual illusions of false horizons. Without a perceptible true earth-sky horizon, any roughly horizontal visual boundary separating differences in brightness is perceived as the true horizon and is used by pilots as an attitude reference. Since most false horizons of this kind are lower than the true one (e.g., a water-land boundary or the lights of a city against a dark surrounding terrain), when the pilot cues on it, he des- cends even wh lie intend ing to mat ntai n level f 1 ight. Another nighttime loss is the result of illumination from instru- ments inside the cockpit or city lights outside. A dark-adapted pilot,
188 flying in ambient scotopic light environments, such as a moonless but starry night, suffers sensitivity losses whenever veiling illumination also enters his eyes. If the veiling illumination is very intense, he may even become light adapted, resulting in a substantial loss in light sensitivity for a number of minutes. But even at lower veiling inten- sities, the glare from a veiling illumination source produces severe losses. Somewhat related to losses produced by veiling light, there is evi- dence that extremely intense illumination over a long period of time may exert a long-term depressive effect on subsequent scotoE,ic visual functioning (Clarke, 19711. This would be relevant to relatively long- duration flight with transitions from daytime to nighttime. Even sun- glasses and visors do not drop the intensity of direct sunlight very much in the exposed cockpit, so that subsequent night operations may be compromised. Another loss occurs with respect to visual search capabilities. Because cone functioning effectively ceases below twilight light levels, the fovea becomes a functional blind spot at night, so that directing one's gaze to the place where one might expect to find sDme- thing is the wrong strategy. This means that well-established search patterns that are most effective during daylight viewing must be inhi- bited and replaced by nonintuitive ones during nighttime. Pilots must learn to look off to the side of a potential target to increase their chances of detecting its presence, and even then detection is only in terms of light discrimination. One of the ways to describe changes in search capabilities from day to night is in terms of a shift from nwhat" determination to nwhere" determination, a distinction that also underlies central-peripheral visual field differences (Liebowitz and Post, 1982~. Recent evidence showing that it is possible to attend to locations in the visual field other than the location being fixated (e.g., Posner, 1982) suggests that it should be possible to develop training programs that stress expectations of what would be seen as a technique to increase detection of targets at night. A problem related to visual search concerns night myopia--a failure to maintain distance accommodation at night (Owens, 1984~. Presumably, because of fewer target objects on which to fixate, the accommodative index of the lenses drifts inward to only a few feet, even when the pilot is actively searching distant space. This reduces the quality of the retinal image patterns, further lessening the pilot's ability to detect targets or even lights at night. Two additional problems, both related to oxygen levels, produce even further difficulties for the pilot flying at night. Hypoxia induced by altitude reduces receptor and neural functioning in the retina (especially of the rods), resulting in substantial sensitivity losses. Most relevant for this context, Kobrick et al. (1984) showed that rod dark adaptation functions were abnormal--with lowered rod sen- sitivity and delayed recovery in the dark even at 14,000 feet. While the effects became progressively more pronounced over a period of days, losses were detectable within the first day at these altitudes. Higher altitudes, without supplemental oxygen in sufficient supply, produce even larger losses.
189 The other oxygen-related effect on night vision arises from accel- eration forces that interfere with normal blood flow to the retinas. Tipton et al. (1984) showed substantial visual acuity losses at meso- pic luminance ranges (presumably, primarily cone functions) at even an acceleration of +2 G. with rapid onset times of the losses. While these researchers did not test their subjects under scotopic conditions as well, it is reasonable to expect that the losses during night flying are much greater than those in daylight. Even minimal maneuvers in mil- itary fighter aircraft are likely to generate sufficient gravitational forces to have an impact on night vision. Because of the massive reduction of ambient illumination at night, light sources that had been invisible during the day are sufficient to be detected at night. The two most important of these, for the pilot, are stars in the sky and lights on the ground. Except for internal instrument lights, stars and ground lights are usually the only visible objects at night. However, because of their size and the absence of any visible terrain, both of these simply act as point sources. ~ have already commented on the possibility that ground lights might be able to provide terrain contour information through depth-fromrmotion processing. Stars cannot do this, and in addition, they provide an impossibly complex background against which to detect a light that is moving very slowly, such as a distant lighted aircraft or one that is not moving at all relative to the pilot's vision (such as another air- craft on a collision course or a tower in the flight path). However, for lighted objects whose angular displacement across the pilot's field of view is fairly rapid, stars provide a frame of stationary lights that make it easier to detect the lights of another plane moving laterally. Even a very starry night does not help much with defining a horizon. Atmospheric attenuation usually reduces the contrast of the stars near the horizon below detectability, which blurs the horizon. Since ground lights near the horizon are also attenuated, no contrast can be per- ceived between the absolutely motionless stars and the relatively motionless ground lights near the horizon. In the absence of a clear horizon, and with point sources both above and below the true horizon, pilots often perceive a false horizon defined by some other luminance di scant inu i ty. In summary, nearly every way in which we perceive the layout of space and our movement through space changes as the lights go out. Furthermore, most of these changes are losses of visual information, dramatically reducing the ability to perceive the scene and one's changing position in it. Liebowitz ~ in this volume) only sightly exaggerates when he says that there is no space perception at night. There is some, but it is minimal, of ten misleading, and usually d is · - Or tent 1ng . SCREENING FOR INDIVIDUAL DIFFE=NCES IN NIGHT VISUAL PERFO~NCE All of the visual processing difficulties at night described above are based on evidence collected from normal observers. Research studies rarely provide very much information about variations from one observer
190 to another--in fact, most studies use only a small number of subjects who are carefully screened and highly trained, so that individual dif- ferences are minimized. The massive testing programs initiated by the military during and after World War II provided ample evidence of variation among the per- sonnel being tested. However, as noted by Johnson (1985), it is not possible to interpret most of that variation since few of the tests had good reliability, and even fewer provided evidence of any validity. Consequently, little is known about whether any of the visual responses to spatial or intensive stimulation presented under scotopic conditions show significant individual differences in a population of potential or actual pilots. One variable has produced reliable and large individual differences in night vision performance that might be relevant to military flying: the age of the observer. While most visual functions show losses with age, this is particularly true for those tested under scotopic condi- tions. For example, Sturgis and Osgood (1982) showed that visual acu- ity and contrast sensitivity were substantially and monotonically lower for older as compared with younger subjects (ages ranged from 20 to 6G) under both photopic and scotopic conditions, but that the losses were larger when testing night viewing conditions. Furthermore, older sub- jects were more affected by glare, reducing their sensitivity even fur- ther. These effects have been reproduced in a number of studies. It should be stressed that such effects cannot be accounted for by the pre- sence of visual pathologies in the older subjects, since all of them were screened initially and could typical meet photopic visual criteria While not all of the factors are understood, general visual functioning decreases with age, and the functioning that underlies night vision is most affected by these decreases. This means that many older pilots, even those able to achieve satisfactory scores on photopically adminis- tered visual tests, might show significant impairments of their night vision capabilities if they were tested for them. Much of the knowledge of impaired night vision functioning has come from the study of patients with visual pathology. The largest single category contains those with dysfunctions at fecting rods, since any im- pairment of rod functioning would have a heavy impact on night vision. Ripps (1981) has reviewed several kinds of rod dysfunctions, all of which produce night vision problems, rang ing f rom slowed dark adapta- tion, to incomplete adaptation, to no adaptation at all. Other prob- lems concern visual field defects, especially loss of peripheral vision at night. The particular kind of problem depends on whether the dys- function is due to reduced photon absorption by the photopigment (for which vitamin A is often an effective therapy), reduced photoelectri- cal effects from the absorption, abnormalities in the rods themselves, or abnormal neural effects at the rod-bipolar synapse or beyond. Of course, in extreme cases, the pathology is apparent during photopic vision as well and pervades all visual functioning. However, the clin- ical literature is replete with examples of patients with substantial pathologies in their night vision and, yet, who appear to be perfectly normal dur ing daytime visual performance. This means that night vision capabilities cannot be assessed by daytime vision tests.
191 Johnson's review (1985) of previous testing and screening work on night vision provides a sobering perspective of the difficulties of developing an adequate testing program. He shows that much of the pre- vious failures can be traced back to a lack of clarity of what is meant by night vision and what the screening should predict about visual per- formance at night. Any new proposals for screening of pilots for their visual abilities during night flight must define night vision in the context of the relevant performance--flying at night. One direction that must be pursued in the development of screening procedures is to measure all of the underlying scotopic visual func- tions {e.g., sensitivity to light, recovery rate of dark adaptation, visual field changes, residual fine-detail acuity, and residual con- trast sensitivity) in a population comparable with applicants for flight training (see the upper part of Table 7~. Those measures that show significant individual differences can then be refined for use in a screening test. Measures that do not show variation can be ignored. But this approach is insufficient: it does not describe how these underlying visual functions relate to actual visual performance during night flight. How well does the applicant perform at night; how well can he see the horizon on a dark night; how sensitive is he to the relative motion of point sources of light on the ground in order to perceive the contours of the terrain or the presence of obstacles in his flight path; how good is he at perceiving and maintaining the cor- rect glide slope, aircraft attitude, and alignment to the runway during the course of landing at night; how good is he at detecting the presence of another aircraft at night or maintaining formation with a wingman who is barely visible? I have added a list of these variables to the lower half of Table 7. To find screening tests that predict these nighttime visual perfor- mance tasks, there must be a research program that develops both kinds of measures--simple tests of visual functioning and realistic measures of flying performance at night. When these two sets of measures are administered to the same group of subjects, then and only then can the components of night vision functioning that affect night flying begin to be determined. The visual functioning tests should include both examples of night vision performance (e.g., dark adaptation functions, visual fields) as well as some that sample those day vision capabili- ties that might be correlated with night vision. Many of the night flying performance measures can be assessed in a relatively simple computer-g~nerated visual imagery flight simulator. While issues of the validity of performance assessment in flight simu- lators are relevant, there already is substantial positive evidence of the validity of the relatively simple assessment procedures. Such gen- eralization is also likely to be acceptable, given the similarity of the visual display of a simulated night scene and the real world at night as viewed from an actual aircraft during night flight. This is because virtually the entire source of stimulation to a pilot at night comes from point sources of ground lights and stars, plus his lighted instruments, and these are relatively easy to display in a simulator. A number of night flying tasks can be assessed with a pilot flying such a simulator. One assessment might be to judge the location of the
192 TABLE 7 Some Possible Variables for Night Vision Screening Tests Laboratory Task Variables Light sensitivity Dark adaptation recovery rate Visual field size and irregularity Fine detail acuity Contrast sensitivity Flying Task Variables Locating horizon at night Extracting contour and obstacles from movement of point sources of ground lights Perceiving glideslope Perceiving alignment to runway Perceiving convergence with runway for proper flare Seeing other aircraft Maintaining formation with friendly aircraft Detecting lighted targets (surface and air) Estimating AGL horizon at different simulated altitudes (in the absence of the appro- priate round gauges and head-up display indicators) and different atti- tudes of his aircraft. Another assessment could be to judge the contour shapes of the terrain or to detect particular terrain elevation features from the shape-fromrmotion information available on the display. Ano- ther assessment would be to detect lighted moving objects seen against a mosaic of stars or ground lights, when the luminance of the background lights and the target varies, as well as the speed of the target across the background. Target detection of objects darker than the background, such as the presence of other aircraft, can also be tested. The various visually guided tasks that comprise landing can be assessed similarly in the same simulator. Each of these different tasks can be tested at mes- opic and scotopic adaptation levels, after varying periods of time in the dark and even under varying degrees of hypoxia. A number of critical questions can be answered from these kinds of data. Most importantly, how much variation is found in the night flying visual tasks among subjects equivalent to applicants to flight training school? This defines the magnitude of the problem facing the development of screening tests. I assume that the variation is signi- ficant, so that it is important to screen out some otherwise qualified applicants because of their poor night vision. A second question, one that has impeded past ef forts to develop screening tests, concerns the relationsh ips among the various night flying tasks that depend on vision. Are they all related to each other, so that knowing how well a pilot performs on night landings will predict how well he performs on all other night vision tasks? It is likely that they are not all related. Rather, there will be several independent components of
193 visual performance needed to fly well at night, and a unidimensional description of night flying will never be sufficient. Furthermore, a unidimensional screening test will not be adequate since it necessar- ily only predicts performance on some but not all of the night flying tasks. Once something about the factor structure of night flying tasks is known, then screening tests that are sensitive to each of the compo- nents can be developed. This will also allow discovery of the reasons for the individual differences. What must be strived for is a subset of tests, laboratory or simulator, scotopic or photopic, that can be devised to predict all of the performance tasks spanning the entire range of night flying. Until there are answers to questions like these, we cannot afford to develop laboratory screening tests whose validity in predicting visual flying tasks at night is unknown. Nor can we begin with any assumptions about how unitary or multidimensional is a night vision flying ability. The only way to find out is to test all components and then empirically determine their relationships and their dimensionality. IMPLICATIONS FOR THE TRAINING OF NIGHT VISION ABILITIES AND. SKILLS This section is necessarily speculative and will remain so until the results of the research outlined in the previous section are known. Even so, there are several aspects of visual performance during night flight that might be amenable to practice and training. As far as I know, little research has been done on any of these. One exception is nighttime target search. For example, Johnson (1985) comments that British rear turret gunners flying in night bombers were trained to distinguish between the shapes of friendly bombers from hostile fighters when viewed at night. The training concerned atten- tion to the outline profile of other planes seen in the dark. In modern terms this training would be described as learning to discriminate the low-spatial-frequency profiles of the two types of targets. I assume that there have been other instances of nighttime target recognition training, all of a somewhat intuitive sort. It is possible to go well beyond intuition, however, with the help of some data. In addition to systematic changes in the appearance of targets as the light levels drop, changes in the perception of range, layout, contours, size, and shape of objects and terrain as the light levels shif t from photopic to scotopic need to be known. As one exam- ple, it is already known that pilots tend to flare on final approach to landing a bit too high at night, suggesting that there are problems in estimating their distance to the ground. Once the nature of any percep- tions that are altered after dark is known, pilots can be trained spe- cifically to correct their behavior accordingly, and this is built into most nighttime landing training programs. Besides nighttime visual search and night landings, the other tasks involved in night flying are rarely mentioned in the training litera- ture, though any of the tasks might benef it f rom practice. For exams pie, can a pilot become better at perceiving the contours of the terrain below him from practicing and observing appropriately moving point light
194 sources in a simulator? Can he learn how to search for targets at night more effectively by practicing at off-center fixation? Can he learn to overcome spatial disorientation better by practicing at detecting a nighttime horizon or by learning to ignore vestibular inputs when there are no corresponding visual ones? Can he learn to overcame night myopia by throwing his accommodation farther out? Can he learn when to expect false horizons and thereby avoid the potentially lethal horizon illu- sions at night? Can he learn to minimize the effects of glare, either by learning where not to look or when not to look? Can he learn to adjust to the systematic distortions of size, distance, layout, contour, and shape that have been noted at night? And so forth. There are no secrets about the kinds of items that should be on this list of training questions; however, the work needs to be undertaken. In conclusion, I have presented an explicit task analysis of the visual demands on military pilots for the accurate perception of space during day and night flight. This analysis permits identification of each instance in which visual information is lacking or visual proces- sing is altered for the particular flying task involving space percep- tion that must be performed at night. From this base, I then examined how to proceed in developing screening tests for night vision abilities and how to address questions presented by training pilots on how to improve their nighttime flying performance thy training their relevant visual abilities. REFERENCES Braunstein, M.L. 1975 Depth Perception Through Motion. New York: Academic Press. Clarke, B.A.J. 1971 Sunlight and night vision. Journal of the British Astray nonlocal Society 81: 208- 210 . G insbu rg, A . 1981 Spatial f iltering and vision: Implications for normal and abnormal vision. Pp. 70-106 in L.M. Proenza, J.M. Enoch, and A. Jampolsky, eds., Clinical Applications of Visual Psycho- physics. Cambr idge: Cambridge Un iver s ity Pres s . Haber, R.N. 1985a Toward a theory of perceived spatial layout of scenes. Computer Vision, Graphics, and Image Processing 6:282-321. 1985b The control of mobility by perceived spatial layout: Application from sighted to blind travelers. Pp. 431-461 in E.R. Strelow and D. Warren, eds., Visual Prosthetic Devices for the Blind. Amsterdam: Martinus Nijhoff. Hecht, S. 1928 The relation between visual acuity and illumination. Journal of General Physiology 11:255-281. Johnson, P. 1985 Testing and train) ng . In P . Johnson, Night Vision, unpub- lished manuscript.
195 Kobrick, J.L., H. Zwick, C.E. Witt, and J.A. Devine 1984 Effects of extended hypoxia on night vision. Aviation Space and Environmental Medicine 55:191-195. Krok, R., and D. Regan 1984 Visual test results compared with flying performance in tele- metry-tracked aircraft. Aviation, Space, and Environmental Medicine 54:906-911. Leibowitz, H., and R.B. Post 1982 The two modes of processing concept and some implications. Pp. 343-364 in J. Beck, ea., Organization and Representation in Perception. Hillsdale, N.J.: Lawrence Erlbaum Associates. Marr, D. 1982 Vision. San Francisco: Freeman Press. Miller, M. 1983 Manual for Low Altitude Flight. Tucson: Arizona Air National Guard. Owen s, D .A . 1984 The resting state of the eyes. American Scientist 72 :378-387. Posner, M.I. 1980 Orienting of attention. Quarterly Journal of Exper imental Psychology 32: 3-25. ., and K.I . Beverly Visually gu ided locomotion: Psychophysical evidence for a neural mechanism sensitive to flow patterns. Science 219:727-732. Persson, L., H. Leyon, and H.~<Marmolin 1982 Factor analytic description of night vision tests. Foersvarets Forskinganstalt U-8310 (August) :1-24 (Stockholm). Regan, 1979 Ripps, H . 1981 Rods, rhodopsin, and the visual process. Pp. 152-170 in L.M. Proenza, J.M. Enoch, and A. Jampolsky, eds., Clinical Applications of Visual Psychophysics. Cambridge: Cambridge Univers ity Press. Sturgis, S.P., and D.J. Osgood 1982 Effects of glare and background luminance on visual acuity and contrast sensitivity: Implications for driver night vision testing. Human Factors 24: 347-360. Tipton, J.L., A.R. Marko, and D.A. Ratino 1984 The effects of acceleration forces on night vision. Aviation, Space, and Environmental Medicine 5S :186-190. Ullman, S. 1979 The Interpretation of Visual Motion. Cambr idge, Mass.: MIT Press. Wallach, H., and D.N. McConnell 1953 The kinetic depth ef feet. Journal of Exper imental PsYchology 45:205-217. Warren, R., and D. Owen 1982 Functional optical invariants: A new methodology for aviation research. Aviation, Space, and Environmental Medicine 53 :977-983.
RECENT INVESTIGATIONS OF GLARE AS A FACTOR IN VISIBILITY AT NIGHT H. Richard Blackwell and O. Mortenson Blackwell The term glare is widely used by laymen and scientists, usually without presentation of a rigorous definition. It usually refers to situations involving the presence of a bright light in an otherwise rather dark environment, the report being made by an observer experi- encing either visual discomfort or visual disability or both. Here glare is defined as producing visual disability with or without visual discomfort, and the experience is attributed to the presence of light scatter within ocular components, of which the cornea is the most important. BACKGROUND The classical work in the field of disability glare dates back to the work of Holladay (1927) and of Stiles (1929~. These investigators presented the basis for the now widely accepted explanation that disa- bility glare is entirely due to light scatter. The evidence of Holladay and Stiles depended on the functional characteristics of the disability glare phenomenon, the loss in visual sensitivity being dependent on lin- early additive effects from different localized glare sources, and the disability being describable in terms of the concept of an equivalent veiling luminance. The basic disability glare equation that emerged from the work of Holladay and Stiles may be rewritten for convenience in the following form: L K ~ s cos v ~2 (1) where Lv is the equivalent veiling luminance in cd/m2; Ls is the lumi- nance of an individual glare element of the task surround in cd/m2; ~ is the angle between the glare element and the task detail in degrees; ~ is Stanley Smith, Ohio State University, presented this paper on behalf of the authors, who were unable to attend. 196
197 the area of the glare element in steradians; and the constant K is the disability mare constant, a proportionality constant that expresses the degree to which the eye of an observer produces scattered light per unit of glare element luminance. The summation in Equation 1 is taken from 8= 1 degree to 8= 90 degrees. The numerical value of K is well established as approximately 10, at least for observers in the 20- to 30-year age group and equilibrium adaptation to about 100 cd/m of luminance. K AS A FUNCTION OF LUMINANCE AND OBSERVER AGE The results of recent studies concerned with possible systematic effects of luminance, observer age, or both on the value of the disa- bility glare constant can be found in a publication put out by the International Commission on Illumination (CIE, 1981~. Clear evidence is presented that both luminance and observer age influence the value of K, with the glare constant increasing as either luminance is reduced or age is increased. More recently, we collated data from several sources and report here for the first time average analytic functions for relating the value of K to luminance, observer age, or both. The analytic functions are as follows: K20 = 8.0 ~ 3.523 (p - 1.82), (2) where K20 is the component of K related to luminance, with age fixed at 20 years and p is the diameter of the ocular pupil in millimeters. Results of previous studies (CIE, 1981) provide the experimental basis for Equation 2, as shown in Figure 1. Equation 2 can be used to relate K20 to luminance by means of the relationship shown in Equation 3, which is derived from the widely accepted work of de Groot and Gebhard (1952~. p = antilog [0.8558 - 0.000401 (log L + 8.603, `3) where p is pupil diameter in millimeters and L is luminance in cd/m2. The relationship between K and observer age was derived from the data collation shown in Figure 2. Our previous studies provided the data point for the two higher luminances; the data for the lowest luminance was made available by Werner Adrian (personal communication). Equation ~ summarizes the analytic functions relating K to observer age, based on data from more than 2,100 observers. A < 42.8 Krel = 1 (4a) A > 42.8 Fret = antilog [1.778~1Og A - 1.6311, (4b) where A is observer age in years. Finally, the joint effects of lumi- nance and observer age on K can be calculated from Equation 5: K = K20 x Krel (5)
198 20 a y 15 10 : 0 1 2 / / K~o. a.o ~ 3.523 (p- 1.82) 3 ~ 5 Pupil Diameter, p (mm.} . ~ 6 7 8 FIGURE 1 Changes in K20, the disability glare constant for 20- to 30-year-olds, as a function of pupil diameter in millimeters. 2.5 2.0 - o J I.S y 1.25 1.0 0 / Kre' ~ onti log [ 1.7 78 ( 109 A - 1.631 )] : o 1 1 I _ 1 1 ~I . ° 100 cd~m2 1.7 cdimt 0.1 cd Zmt 1 D 15 20 25 30 40 SO 60 70 80 100 Age (Years,Log Scale) FIGURE 2 Changes in Krel' the disability glare constant set to unity for 20- to 30-year-olds, as a function of observer age.
199 CALCULATION OF DISABILITY GLARE EFFECTS UPON TARGET VISIBILITY Blackwell (1955) first pointed out that the addition of ocular stray light due to glare sources always produces the deleterious effect of a reduction in image contrast and may or may not produce a potentially beneficial effect due to adaptation to a higher luminance. In dis- cussing the calculation of disability glare effects on target visi- bility, these effects should be kept separate. Consider, first, what may well be the most usual situation of inter- est, in which the effect of disability glare is expressed relative to a situation of zero stray light. Then, the effect of disability glare in reducing image contrast is expressed in terms of the contrast factor (CF), CF = L/(L ~ Lv), (6) where L is the actual background luminance and LO is given by Equa- tion 1. It also sometimes occurs that the effect of disability glare is expressed relative to a situation of a perfectly uniform extended field which, without any glare sources as usually defined, produces 7.4 percent stray light. In this case, the effect of disability glare in reducing image contrast is expressed in terms of the following: CF = L/Le, where Le is Le = (L + LV)/1.074. em. (7) (8) In either case, disability glare reduces target visibility by reducing image contrast, and the magnitude of the effect differing between the two referencing systems. The potentially beneficial effect of disability glare arising from the increase in contrast sensitivity to be expected from adaptation to a higher overall luminance only occurs if adaptation occurs during the time span of interest following the introduction of ocular stray light. Assuming that complete adaptation has occurred, a readaptation factor (RF) can be computed which is defined as follows for the situations in which the zero stray light baseline is used: RF = RCS for (L + LV)/RCS for L, (9) where RCS is relative contrast sensitivity and is derived from the rela- tionship given previously (CIE, 1981~: RCS = 1.555 1.639 0 4 + 1 -2. 5 [( L ) 2 ] where L is the luminance ( in cd/Tn ) to which the eye is fully adapted. For the situations in which the 7.4 percent stray light baseline is used, the Readaptation Factor is as follows: RF = FCS for Le/RCS for L. (10) 11)
200 Whichever baseline is used, the overall effect of ocular stray light on target visibility can be computed by means of the disability glare factor (DGF): DGF = CF x RF, (12) provided it is assumed that full adaptation to the higher luminance produced by ocular stray light has occurred. Comparison of the values of DGF and CF reveals the extent to which losses in target visibility due to reduction in image contrast are offset by increases in target visibility due to increased contrast sensitivity resulting from adap- tation to a higher luminance. ASSESSMENT OF Lv IN NIGHT ENVIRONMENTS OF INTEREST Equation 1 implies that assessment of LO in a night environment of interest requires the measurement of luminance and angular size of each elementary source of disability glare, followed by weighted summations to obtain the total glare effect of all sources in the visual field. Although such a procedure is theoretically possible to follow, it is not generally considered a practical way to proceed. Rather, there is generally widespread use of an optical attachment to a luminance pho- tometer. The attachment is designed to provide the summation of Equa- tion 1 automatically. The optical attachment consists of a transparent block of Lucite plastic, into which an aspheric crater has been preci- sion formed. The crater is cut into the block, with the aspheric sur- face facing forward with respect to the objective lens of the photo- meter, there being a piano surface of the attachment facing away from the objective lens. Light rays directed toward the photometer must pass through the clear Lucite block and are refracted when they leave the Lucite block. Rays reaching the Lucite block from a given angle from perpendicular to the piano surface of the block are refracted suitably so that they reach the objective lens as if they originated from a point straight ahead of the photometer through only a narrow annular region of the aspheric surface. Thus, the width of the annular region of the aspheric surface that refracts rays from a particular point in space determines the light flux that is refracted. According- ly, the specific shape of the aspheric surface defines flux weighting factors for rays coming from a particular point in space so that they are refracted into the photometer acceptance cone. Indeed, the aspheric surface contour defines relative flux weighting factors for rays coming from all particular points in space. The precise contour of the aspheric surface required to weight light flux coming from different angles 8, so that they are in accord with the function described in Equation 1, was determined by operating the calculational sequences in reverse. Then, a "male mold" was cut by precision machining techniques, as verified by a shadowgraph. Finally, the male mold was heated and used to form the aspheric contour in the Lucite block. The success of the calculations and machining operations was verified by measuring the responses of the photometer to single rays originating at first one and
201 then another value of 8. Typical results are shown in Figure 3. me solid curve represents values of Lo, normalized to the photometer res- ponse at ~ equal to 1 degree, and multiplied by 100. The X's represent experimental values of relative LO as a function of ~ for a specific member of the optical attachment family. me agreement between the X's and the solid curve is considered remarkably good, considering that values of relative Lv varied over nearly 5 log units. It should be apparent that the optical attachment, often referred to as a glare lens, is well qualified to mimic the ocular stray light characteristics of the 20-year old eye when operating at relatively high luminances. Values of Lv obtained with such a glare lens attachment to a luminance photometer can be adjusted to take account of the effects of luminance and observer age on the value of Lv. Thus, glare lens photometric attachments seem to offer a very useful means of determining values of Lv in night environments of interest. The interested reader should refer to the paper by Fry et al. (1963) for a more detailed description of the procedures used to define the contour of the aspheric surface and its male mold. However, the paper by Fry et al. (1963) creates a problem for the careful reader since the basic equation used to describe the functional characteristics of the Lv function of ~ differs in an important way from equation (1) of the present paper. me form of Equation 1 follows the recommendation by Pry in earlier work which leads to the revised Eguation 1 presented as follows: L = K ~ s cos ~ 8~8 + 1.5) (Fry et al. revision) (13) where all quantities have the same meaning as in our original Equa- tion 1. In this case, the summation is taken from ~ = 0 degree to ~ = 90 degrees. Indeed, one of the obvious advantages oF the Fry et al. revision of Equation 1 is that the summation has rational rather than arbitrary summation limits. Fry et al have also argued that the revised form of Equation 1 conforms better to theoretical expressions for light scatter in turbid media than does the original form. Perhaps for these reasons, the Illuminating Engineering Society of North America has in past years chosen the Fry et al. revision rather than the classi- cal Holladay-Stiles formula. On the other hand, the International Com- mission on Illumination has in effect chosen the classical formula in its promulgation of Publications 19/2.1 and 19/2.2 (CIE, 1981~. Fortunately, the problem of choosing between Equation 1 and 13 may be left with the visual scientist or illuminating engineer so far as use of the glare lens optical attachment is concerned. The basis for the freedom of choice that is left to these users of the glare lens may be judged by reference to Figure 4. Here the solid curve values of Lv are presented, normalized to the photometer response at ~ = 0 degrees, and multiplied by 100, as specified by Eguation 13. The X's represent experimental values of relative Lv as a function of ~ for the very glare lens for which the data were presented in Figure 3. In fact, the two sets of X's differ only with respect to their arbitrary placement
202 Conformity of 1983 Glare Lenses to Formulation loo ~ 50 _ 10 5 S._ . - 1 1 .05 .01 .005 .001 1 1 1 1 o 1 , , \ Lv = Cos B/ 82 it. - ~I ~r ~I . I · l l X 80 FIGUPE 3 Degree of conformity of measured values of veiling luminance, Lv' as a function of glare angle, 8, to curve prediction by Equation 1.
203 on the graph ordinates in the two cases. file placement is arbitrary in the sense that the data set may be moved up or down on the logarithmic ordinate to achieve best fit to the solid curve in each case. Of course, the calibrated glare lens constant used to specify the photo- meter value of Lo depends on the up and down placements in the two cases. In other words, the same glare lens has two different calibra- tion constants when considered to represent Equation 1 or 13. However, the important point here is that, at least to the observant eye, the goodness of fit of the X's to the solid curve is approximately the same for the two formulations. This seems to signify that the same glare lens can be used to represent Equation 1 or 13 as desired, with the calibration constant used in determining the value of Lv in a night environment of interest being different in the two cases. Accordingly, an experimentalist could readily ascertain the extent of difference in the calculated value of Lo to be expected from Equation 1 or 13. The form of Equation 1 and 13 suggests that the two values Of Lv differ substantially only when most glare light reaches the glare lens from angles of ~ of less than about 10 degrees. It is our experience that small angle glare of this sort is relatively rare in many night environ- ments, so that it is perhaps not a crucial decision that must be made in selecting between Equation 1 and 13. DI SABI LI TY GLARE AS A MAJOR FACTOR IN NIGHT VISION AND TARGET VI SIBILITY ,. The loss in image contrast due to ocular light scatter in the pre- sence of glare sources often dominates the assessment of night vision capability and target visibility. The readaptation effect can provide substantial amelioration of the visibility loss, provided that the night vision situation permits the observer time to readapt fully to the increased luminance due to light scatter. However, in many crucial night vision situations, it is the initial moment of exposure to the glare sources that creates the most significant visibility problem, and during this moment there cannot be any readaptation so that the image contrast loss must be accepted without amelioration. Any interested party can assess this conclusion by initiating empirical measures of Lv in night environments of interest. me same interested party should also have some practical experience in measuring the value of the disability glare constant for a variety of observers differing in age and with different luminances of the night environment. Our work has convinced us of the efficacy of our special method for measuring K. We measured visual detection probability as a function of target contrast for two situations put in direct comparison. In one case, there was an extended background and a surround of uniform luminance. In the other case, there was a bright annulus of glare light with an inner diameter of 3 degrees and an outer diameter of one or ano- ther size, depending on how large a contrasting loss was desired. The essential characteristic of the annulus is that the annulus luminance should not exceed the backg round luminance in the compar ison exper iment by more than about 25 times. Under these conditions, the observers can
lo 0 ~ - ~ is - .01 1 . . 1 _ .005 .001 . 1 ~ \ \ \ \ \ \ \v~ 204 Conformity of 1983 Glare Lenses to Formulation my = Cos 9/~8x(~+l,51] . 1,i, 0 10 20 30 40 l \x Ax \ x 1 50 60 70 80 FIGURE 4 Degree of conformity of measured values of veiling luminance, Lv, as a function of glare angle, 8, to curve predicted by Equation 13.
205 resist the temptation to glance occasionally at the bright annulus and thus can allow an accurate assessment of disability glare (rather than an unwanted assessment of transient adaptation). Our experience shows that in the classical experimental paradigm involving a single small bright glare source, the observer's temptation to glance occasionally at the bright glare source cannot be avoided; this contaminates the experimental data by including at least some transient adaptive effect. Our glare annulus tests were administered at either a moderately high or a moderately low background luminance without encountering this difficulty. REFERENCES Blackwell, H.R. 1955 The use of visual brightness discrimination data in illu- minating engineering. Compte Rendu, 13th Session of the CIE, Sol. I, Sec. 3.1.1.1. de Groot, S .G ., and J .W . Gebha rd 1952 Pupil size as determined by adapting luminance. Journal of the Optical Society of America 42:492. Fry, G.A., B.S. Pritchard, and H.R. Blackwell 1963 Design and calibration of a disability glare lens. Illuminating Engineering 58 :120-123. Holladay, L .L . 1927 Action of a light source in the field of view in lowering visibility. Journal of the Optical Society of America 14 l-a . International Commission on Illumination (CIE) 1981 An Analytic Model for Describing the Influence of Linhtino _ Parameters Upon Visual Performance, Vol. 1: Technical Foundations; Vol. II: Summary and Application Guidelines. Publications CIE No. 19/2.1 and 19./2 . 2, Par is: International Commission on Illumination. (Available through the U.S. National Committee, CIE, National Bureau of Standards, Washington, D.C.) Stiles, W.S. 1929 The effect of glare on the brightness difference threshold. Proceedings of the Royal Society (London) B104:322-350.
DUPLIC ITY, PHASE LAGS, AND DESTRUCTIVE IN=R"~NCE IN TOPIC AD ==OPIC FLICKER PERCEPTION Donald I.A. MacLeod and Andrew Stockman Recently there has been a lot of interesting work on how the dynam- ics signals change with changing light levels and with the transition from cones to rods. These changes are reflected in our capacity to produce and analyze both flicker and motion. In this paper we concen- trate on flicker because it is simpler to produce and analyze and (pre- sumably as a result) has been studied in more detail. The traditional way to characterize visual temporal resolution is to determine the highest frequency of flicker that can be seen. This frequency, above which a flickering light appears steady, is referred to as the critical flicker fusion frequency or CFF. CFF varies with intensity in a characteristic way, which has been produced in thousands of experiments, notably those of Hecht and Shlaer (1936~. Generally, the more intense the light the higher the frequency that can be re- solved. m is improvement, however, occurs in two well-defined phases. First, the rods mediate flicker detection. They typically reach a limit at about 15 Hz, so that further increases of intensity produce no improvement in the CFF. Then, the cones take over detection and allow a further increase up to above 50 Hz where they reach their limit. Unfortunately, such data indicate little about the dynamics of the visual system, that is, about the time course of the visual signals. For instance, the rise in CFF with increasing intensity might occur because the visual signals persist for a longer time in dim light. But, alternatively, it could be explained--and this was essentially how the model of Hecht and Shlaer (1936) explained it--by an improvement in differential sensitivity in bright light that allows detection of a smaller modulation or a smaller functional change in excitation. In the 1950s, however, DeLange (1952) established a method of character- izing the dynamics of the visual system much more completely--through its f requency response. To measure the visual f requency response, the time-averaged intensity is kept constant. For the data shown in Research supported by g rant EY 01711 f rom the National Institutes of Health and a Postdoctoral Fellowship from the North Atlantic Treaty Assocation awarded to A. Stockman. We thank Sandy Bruce for help in running experiments and Kathy Purl for drawing f igures. 206
207 .01 .02 . ~ ~ : .05 .1 .2 · Cones 4000 ph. td. DIAM \ \ . ..< , , . , . . . 40 10 20 frequency ~ Hz) 30 FIGURE 1 Cone modulation sensitivity measured as a function of tem- poral frequency at an adaptation level of 4,000 photopic trolands (ph td). Test field, 668 nm; 6.2-degree diameter; adapting field, deep red (Wratten no. 70 gelatin filter), 11.5-degree diameter; fixation, 13 degrees temporal; observer, DIAM. Figure 1, it was set at a high photopic level (4,000 photopic tro- lands). Then, for each flicker frequency, the observer' s sensitivity is determined by f inding the smallest modulation at that frequency that can be distinguished from a steady light. The smaller the threshold modulation, the greater the sensitivity (in Figure 1 sensitivity is plotted increasing upward). Sensitivity holds up well at frequencies up to 30 Hz and then falls off with increasing rapidity. With rod vision at a low scotopic light level, however, the fre- ouency response looks very different, dropping sharply with increasing frequency until it reaches its limit at about 15 Hz. So rods, unlike cones, are extremely sluggish at low light levels.
208 ROD VI SION IN THE MESOPIC RANGE The first important point to be made here concerns the performance of rod vision in the mesopic range. This is difficult to investigate because in this range the rods are upstaged at high frequencies by the more agile cones. Conner and MacLeod (1977), however, were able to characterize the behavior of rod vision at these intensities by adopting special procedures to keep the cones out of the way. The results were surprising. What we did was superimpose a flickering test light of a middle wavelength (actually green), to which rods are highly sensitive, on a red steady background to which cones are much more sensitive than the rods. For cones the steady red completely swamps the flickering green; for rods the red is almost negligible. In addition, the test light was presented obliquely incident to the retina. This was done because the cones are significantly less sensitive to obliquely inci- dent light than to light that strikes them axially, whereas rod sensi- tivity is nearly independent of the angle of incidence. Under these conditions, Conner and MacLeod were startled to find that the rod CFF, after appearing to reach an asymptote at 15 Hz in the scotopic range, suddenly begins to improve again at levels above 1 troland (td), so that the CFF-log I intensity curve is still double-branched, with a second conelike rod branch that is normally hidden behind the real cone branch (Figure 2~. It is known that this second rod branch is not a cone branch on four grounds. First, CFFs measured during the cone plateau of dark adaptation (filled triangles, Figure 2) require a greater intensity by about a factor of 10 for attaining a given frequency than is needed after rod recovery. Second, if the experiment is repeated with axially incident test lights, this considerably improves sensitivity for cones as shown by the cone pla- teau, but has little effect on either branch of the rod curve (see Figure 1B in Conner and MacLeod, 1977~. Third, when the wavelength of the test light is varied, a rod spectral sensitivity is found on both these branches (open symbols, Figure 2~. Finally, a fourth indicator of the rod origin of the responses is that vision fails, with a preci- pitous decline in OFF, at intensities greater than 100 td, at which it is known that rods saturate. This rod saturation-related decline in CFF, from a peak close to 30 Hz, is illustrated in Figure 6 of Conner (1982~. So there is clear evidence for a duplicity within the rod mechanism itself that allows rods to pick up frequencies at mesopic light levels nearly twice as high as the scotopic limit of 15 Hz found by Hecht and Shlaer (1936~. It has been known for some time that rod monochromats also show a double-branched CFF-log I curve. As one example among many, recent data from a current investigation by Hess and Nordby (in preparation) agree very well with Conner and MacLeod's (1917) data, the only differ- ence being that the low-intensity branch on the rod monochromat starts at a higher detection threshold, as it might if some ambient light were present au-ring testing. The prevailing view on this has been that this second branch in rod monochromats, which does show a rhodopsin spectral sensitivity, is due to rhodopsin-filled cones; this is an unlikely hypo- thesis, especially because the rod saturation above 100 td shows up
28 I 2 4 A, 20 3 v ~ 16 lo i_ in 3 _ 12 ._ . _ lo .001 209 o 430nm /( 469 nm / o 520 nm / a · 469nm (cone plateau ~ 0 . ~ o~ 0 -of lo ~ - -°~°-°'S £8 A/ .01 .1 1 10 100 Luminance ~ Scotopic Trolands) FIGURE 2 Critical flicker frequency measured as a function of stimulus intensity. Settings were obtained either after complete dark adapta- tion and with a test stimulus of 430 nm (open circles), 469 nm (open triangles), or 520 nm (open squares) or during the cone plateau phase of dark adaptation and with a test stimulus of 469 nm (filled triangles). Test field, 9 degrees; adapting field, 670 nm, 13 degrees; fixation, 16 degrees temporal. In this experiment only the test lights (but not the adapting field) were obliquely incident on the retina. The coincidence of the double-branched CFFs (open symbols) supports a scotopic spectral sensitivity for both branches. Data from Conner and MacLeod (1977~. very clearly in rod monochromats as well as in normal trichomats. The evidence for rhodopsin-f illed cones comes f rom an experiment by Alpern et al. (1960) which showed a loss of sensitivity for marginal pupil entry of the stimuli in their observer. However, using the same rod monochromat who was the subject in the Hess and Nordby study, recent experiments by Sharpe and Nordby (1984) have now shown that the sensi- tivity to eccentric stimuli is that of rods not cones. No cone func- tion whatsoever has been found in this subject, so if rhodopsin-filled cones do exist in humans, which now seems doubtful, they are clearly not the basis for the improved mesopic flicker detection in this true rod monochromat, or indeed under our conditions in the normal eye. Instead, the evidence shows that light-adapted rods can detect rapid
210 flicker, at least at f requencies up to 30 Hz. Moreover, the transition between low and high mesopic behavior is extremely abrupt, as if two separate mechanisms are involved. So why is there this abrupt improvement in the temporal resolution of rods in this intensity range? Is it a change in frequency response, or just an abrupt improvement in differential sensitivity? As Conner (1982, Figure 7 therein) has shown, the answer is that the frequency response is dramatically altered. At the lowest scotopic luminance, sensitivity drops off monotonically with increasing frequency. At higher scotopic luminances, still below cone threshold and below the breakpoint found in the rod CFF, the curve begins to take on a bandpass characteristic. This becomes much more pronounced at still higher sco- topic luminance s, until at a level just below rod saturation there is a sharp peak in sensitivity at about 10 Hz. The curve, in fact, becomes quite conelike. However, if Conner's rod frequency response at the highest scotopic luminance that he used is compared with the results for cones at a high photopic luminance, (e.g., Figure 1), cones still come out ahead, with a response much more extended to high frequencies. In this comparison, however, the cones are working at a much higher intensity or quantum flux level than the rods. What would happen it the intensity were set to the same level, in terms of quanta absorbed per second per degrees, for both rods and cones? That comparison is made in Figure 3. The result is that the cones still retain an advan- tage, but it is relatively slight: if rod and cone sensitivities are equated at low flicker frequencies, it is not until 20 Hz that rod sen- sitivity drops by a factor 2, or 0.3 log units, below that of the cones. Therefore, it is found that the bandpass type of temporal response, peaking close to 10 Hz, is a pronounced and general characteristic of visual function under moderate or high illumination. It has not yet, as far as we know, been exploited for practical purposes, but Figure 4 suggests a way in which it might be exploited to improve the detecta- bility of large, low-contrast and indistinct targets, notably when natural vision is degraded by fog or haze. Natural vision relies on haphazard, more or less involuntary eye movements to convert spatial contrast into temporal transients. The optical chopper of Figure 4 does this more efficiently and at a rate close to what is optional for visual sensitivity by presenting the target alternately with a uniform field of equally spaced average luminance, so that any deviations from that luminance across the target region are converted into a highly detectable flicker. The chopper could be used independently or incor- porated into viewing instruments. In electronic image intensifiers it could be implemented more efficiently by making provision for a contrast reversal of the image (rather than simply intermittent presentation) at a visually effective rate. PHASE LAG A relatively neglected but interesting aspect of the frequency response of rods and cones is the phase lag. A phase lag or delay is inevitably found in any sluggish system that has a reduced sensitivity
211 .02 .05 o at_ ~ 1 3 o .' - I_ /\ O' o o ·\ o Cones 25 ph.td. Rods 62.5 sc.td. DIAM A 10 20 30 40 frequency (Hz) FIGURE 3 Modulation sensitivities of the rods (open circles) and cones ~ filled circles) measured as a function of temporal frequency at adap- tation levels of 62.5 scotopic trolands (sc td) and 25 photopic tro- lands (ph td), respectively. The levels were chosen so that the quantal absorptions in the rods and cones are approximately equal in the two cases (see text for details). Test fields, 510 nm (rods) or 668 no (cones), 6.2-degree diameter; adapting field, deep rea (Wratten no. 70), 11.5-degree diameter; fixation, 13 degrees temporal; observer, DIAM.
1 212 \ FAINT TaRGET 6 Hz alternation me\ \ Uniform reference field is equated with target In space average luminance / ~\1 If\\ / A .. \\\ REFERENCE FIELD ( uniform diffuser) -l \\\ Hi\ - - at, ~ / FIGURE 4 An optical chopper to aid low-contrast vision ~ see text for details) .
213 to high-frequency stimuli. The phase lag of rods relative to cones can be measured by stimulating them in alternation, for instance with green and red lights, respectively. Any phase difference in the resulting signals prevents the rod and cone signals from canceling completely when the two signals are added together. (For now the ganglion cell is though to add together the signals from rods and cones.) If the greater delay of the rod signal is compensated for by advancing the rod stimulus by an equal amount, the two signals wild be brought into opposite phase and will cancel to yield a constant sum, that is, a steady or minimally flicker ing light. The amount by which the rod stimulus must be advanced from the opposite phase to minimize flicker is an index of the rod sig- nal's phase lag relative to that of the cones. Figure 5 shows some data on the rod phase lag (relative to that of the cones). The first thing to note here is that the phase data sup- port the idea that rod vision has an internal duplicity, with a fast process operating at moderate mesopic intensities and a sluggish one at low scotopic intensities. The lower curve shows the phase lags recorded at mesopic intensities. At scotopic intensities (the upper curve) the phase lags are considerably larger. At 15 Hz, for example, the SCotQpic lag is nearly a whole cycle, whereas the mesopic lag is a half cycle. The lags in both cases are substantial, except at very low frequencies. They appear to be greater than would be theoretically expected to result from rod visual persistence alone, and they suggest what amounts to an additional latency of close to 50 Ins for the sensitive, slow rod mech- anism and about half that for the insensitive, fast one. So both the mechanisms are quite slow by this criterion. It is almost certainly on account of these phase lags that the technique of flicker photometry has never been successfully applied in mesopic vision. Fortunately, a more recent but related technique may solve the problem. In minimum motion photometry (Anstis and Cavanagh, 1983; Anstis et al., in press), measurements equivalent to flicker pho, tometric ones can be made using drifting gratings. Unlike flicker photometry this can work well at extremely low drift rates, where the phase difference between rod and cone signals appears from Figure 5 to be small. An interesting special situation occurs when the rod phase lag is 180 degrees. Then, a single mesopic flickering stimulus will give rise to rod and cone signals that are in opposite phase, and if these are of the same size they will cancel (Figure 6~. This does happen, and the condition has been called the mesopic null (MacLeod, 1972, 1974~. In the mesopic null, light of a certain frequency looks steady, even though it can be seen to flicker at either higher, cone-dominated or lower, rod-dominated intensities. This is illustrated in Figure 7 for 7.5-Hz flicker, under conditions where a rod-cone phase difference of 180 degrees was actually measured at that frequency. The vertical axis is the intensity of a yellow test stimulus. The filled circles define flicker threshold contours for 7.5 Hz. On an increase in the test in- tensity, a range of intensities is found where flicker is more or less abolished. As would be expected, the addition of a steady background to knock out rods destroys the mesopic null phase and reveals flicker where none was seen before. Nothing like that happens at 4.5 Hz
214 - ROD-CONE PHASE LAGS ° 03s~td. · 62 5 sc. Id. i_ 500 400 300 In - 200 100 o -20 to to' / / O~ / / ·' . ' DIAM - ,. 5 10 15 20 25 frequency ( Hz) FIGURE 5 Phase lags measured at two adaptation levels: 0.3 sc td (open circles) and 62.5 sc to (filled circles). Phase lags were esti- mated by varying the phase difference between a 510-nm flickering light seen by the rods and a 668-nm light seen by the cones. By adjustment the subject found the phase difference at which the resultant flicker appeared to be null, or of least amplitude. Necessarily, the rod phase lags are relative to those of the cones; they are given by the amount that the rod stimulus must be advanced front opposite phase to yield a null. Test fields, 6.2-degree diameter. At the low adaptation level no adapting field was present. The adapting field at the high adap- tation level was deep red (Wratten no. 70), 11.5-degree diameter, and 3,000 ph td. Fixation, 13 degree temporal; observer, DIAM.
215 MESOPIC NULL Single stimulus Cone response - + i\ Rod response -" ALA Summed response Cone Acone " /rod\_/rod\ /rod ~ FIGURE 6 An illustration of the origin of the mesopic null. This type of null can occur under conditions where the rods lag behind the cones by 180 degrees (e.g., at 0.3 sc to and about 7.5 Hz--see Figure 5) . Under such conditions a single mesopic flickering stimulus gives rise to rod and cone responses that are in opposite phases. If the a~.pli- tude of the rod and cone responses are equal, the summed response will be a steady signale Thus, destructive interference between rod and cone signals can cause ~ single stimulus, flickering above rod and cone threshold, to appear s beady . (open triangles) where the measured phase lag under these conditions was only about 90 degrees. The curves fitted to the data are generated by the simplest possible theoretical model (MacLeod, 1974), in which rod and cone signals each satisfying Iceberg law are added together either in opposite phase (for 7.5 Hz) or in quadrature phase (for 4.5 Hz), and flicker visibility depends on the amplitude of the resultant or vector sum. If the rods have an ir,ternal duplicity, a scotopic null where cones are not involved, but where the fast and slow rod signals come out in opposite phase may be expected. This prediction was tested by Conner (1982), and he did indeed discover a scotopic null attributable to roo- rod interference at a frequency of about 15 Hz. To demonstrate this null, Conner found it necessary to present the stimulus as a bipartite field with opposite phase inoculation of the two sides, but more recently we have found it easy to observe the scotopic null without this compli- ~ation (L.T. Sharpe, D.I.A. MacLeod, and A. Stockman, in preparation).
216 ·e \ 1 : ·' I . ·1 lo..! 1 . . ~ ~ of\ ._ o _ o In \ \ it ~ 1 ;~ (Pl) [~!S~alUI M0iDd lSG1 1] ~ . . I= . a' .,. CS a' .= 1 U] :5 _ o A_ I: O o ::S N O ;= Q ' A . ~ ~ · - U] 3 0) a, O ~ ~ 1 ~1 ·. ~ ~ N .,1 ~ Lo On . :D ~ ·~1 a,, us U] C: ~ c: a, a) ~ ~ I: O ~ ~ O I A 00 0 - O AS - - U] o O a) U) O Q4 O · - U) ~ ~ O ~ O4 k4 s :5 U] ~Q ~5 o- - . - a ~3 3 ~ O C: a ~ + ~ o O · - cr, ~ ~ + O ~ ~ O ~ H - U] 0, 0 3 U) ~ ~ ~ C: ~ H . - O a)~< ,. - O ~ ~ O .,' ~ ·~1 a~ ~ ' (: <1) ~ ~ u~ O O a~ ·,' Q' a) Q. U] O ~ S U Q' a ~ ~ Q ~ ~ Ll S O O Q O ~ ~ ·-l t) S 3 ~5 05 ~ ~ ~ s ~5 U] X ~ U] ~ U] U] ~ ~· - ~ Q4 u, a Q. ~ · - 1 a) ·-. ~ tn ~ L' U] U) ~: · - 1 c~ ~ · - s · - 1U] ~i E~ ~S O ~ O £ ~ N $3 H t1S a m~ · _ ~ 0 ~ ~ L4 ~ _ a) c: s ~ 0 ~ ~ ~ · - a) · - a) u, U) ~C SJ ~ Q 3 ~ :) 0 a) 0 ~ ~ X · - a. o co ~r · - ~ ~ ~ U] c: ~ O `; a ~ ^ ~ .,4 ~ N C: ~ ~ :: . - O O U] (L) u~ (L) ~ ~ · 5: ~ - ~£ ~ ~ _ U] O · - ~ 3 ~ a, O O ~ ~ - - ~n - ~: :5 ~ O U] ~ S O :3 o ~5 UD Q - a' ~ U) >1 U] O ~ ~ o c) ~ u ·_1 s ~ c) ~ o o ~ s 3 a U] o ~; V
217 All the standard controls were done--directional sensitivity, spectral sensitivity, comparison with cone plateau--with results indicating that cones are not a factor. The occurrence of a scotopic null at 15 Hz or so is consistent with the measured phase lags shown in Figure 5 for the fast and slow rod mechanisms. However, those lags were measured at very different levels of adaptation, whereas the scotopic null requires the two mechanisms to generate opposite phase signals at one and the same adaptation level. In fact, this requirement is satisfied, because the transition from slow to fast behaviors is abrupt (Sharpe, MacLeod, and Stockman, in preparation). For Conner's interpretation of the scotopic null to be sustained, the net rod signal just on the high intensity side of that null must be in opposite phase with the rod signal from a stimu- lus just on the low side of the null. Sharpe, MacLeod, and Stockman have measured the rod phase lag, relative to that of cones, at stimulus intensities just 0.4 log units apart, straddling the scotopic null. They were found to differ by nearly 180 degrees--strong evidence that the null results from two separate rod signals that destructively inter- fere with each other. The hypothesis that cones are involved is con- tradicted by the observation, in agreement with Figure 5, that on the high-intensity side of the scotopic null the supposed rod signal has a phase lag of nearly 180 degrees relative to that of the cones. Thus, it cannot be a cone signal. So, in addition to the four types of evi- dence already mentioned for the view that the fast and slow processes that yield this null are both rod driven, this result provides a fifth type. THE LIMITS OF ROD SENSITIVITY: DETECTION VERSUS AMPLITUDE DI SCELIMINATION These experiments on rod-cone and rod-rod null phases have lea to one other very striking observation. When a cone stimulus flickers just above the flicker threshold, a rod stimulus of suitably chosen phase can reduce that flicker even when the rod stimulus itself is set invisible (MacLeod, 1974~. This at first sight seems to contradict the simple notion that rod and cone signals just add up, but actually there is no inconsistency--it only needs to be assumed that a high threshold is applied to the resultant (rod and cone) flicker signal so that no flicker at all is registered when the resultant amplitude is below the threshold amplitude. Differences in amplitude above the threshold could then be discriminated with a precision far greater than threshold sensi- tivity measures might suggest. This was tested by measuring interaction or beat sensitivity (using a technique first developed to Study the tem- poral properties of the blue cone mechanism; Stockman and MacLeod, in preparation). To measure the beat sensitivity to a roe stimulus of a particular frequency, it is presented not by itself but along with a flickering cone stimulus that differs very slightly in frequency from the rod stimulus. The cone stimulus remains present at a fixed flicker amplitude (chosen to be close to the flicker threshold) throughout the experiment. The observer's task is to adjust the amplitude of the rod stimulus so that its presence can just be detected in the form of an alternate waxing and waning of the flicker amplitude, as the two
218 ·005 .01 ·02 c 05 .1 .2 S. l\ o 0 Rods ~ beats) · Rods (thresholds) 62.5 so. td. i' o" `, a_ . on o 10 20 30 40 frequency ( Hz) FIGURE 8 Modu let ion sen si t iv i ty ~ f i lied c i rc le s) and bee t sens it iv ity (open circles) measured as a function of temporal f requency at an Adam tation level of 62.5 sc td. For the modulation sensitivity estimates 510-nm test f ields were used. By ad justment the observer set the modu- lation of the 510-nm light at which flicker was no longer seen. For the beat sensitivity estimates, an additional 668-nm test f ield was presented flickering at a slightly higher or lower frequency than the 510-nm light and at a near-threshold amplitude. Under su itable condi- tions this gives rise to an amplitude modulation of flicker at the dif- ference or beat f requency. By ad justment the observer set the mo~ula- tion of the 510-nm light at which beats were no longer seen ~ for details see text) . Test f ields, 6. 2-aegree diameter; adapting f ield, deep red (Wratten no. 70), 11.5-degree diameter; fixation, 13 degrees temporal; observer, DIAM.
219 components of the flicker stimulus come into or out of phase. The fre- quency of this beating was set to be 0.5 Hz or less, and the rod test stimulus frequency was varied from 2 to 40 Hz, keeping the difference in frequency between rod and cone stimuli constant. With this procedure, it was found (Figure 8) that rods can detect intensity modulations as small as 0.5 percent or 0.005 log units, corresponding to a Weber frac- tion of 1/200. That peak sensitivity, which is reached at 10 Hz, is more than 10 times greater than has been achieved in rod vision with a conventional flicker detection task, for which our results shown by the filled symbols in Figure 8 are typical. In conclusion, work on flicker detection by rods has demonstrated three points: (1) there is an internal duplicity of organization, with a sluggish sensitive process and a fast, less sensitive process; (~) the rod signals have large phase lags and can interfere destructively with each other or with cones; and (3) rod modulation sensitivity can be improved by an order of magnitude by providing a fixed stimulus of slightly mistuned frequency for the test stimulus to beat with. REFERENCES Alpern, M., H. Falls, and G. Lee 1960 The enigma of typical total monochromacy. Amer ican Journal of Ophthalmology 50:996-1012. Anstis, S., and P. Cavanagh 1983 A minimum motion technique for judging equiluminance. In J.D. Mollon and L.T. Sharpe, eds., Colour Vision: Physiology and Psychophysics. New York: Academic Press. Anstis, S., P. Cavanagh, D. Mauge, T. Lewis, D.I.A. MacLeod, and G. Mather In Computer-generated screening test for colorblindness. Color press Research Application. Conner, J.D. 1982 The temporal properties of rod vision. Journal of Physiology 332:139-155. - Conner, J.D., and D.I.A. McLeod 1977 - ~ DeLange, 1952 RON photoreceptors detect rapid flicker. Science, New York 195:698-699. H. Experiments on flicker and some calculations on an electrical analogue of the foveal systems. Physica 18:935-950. Hecht, S., and S. Shlaer 1936 Intermittent stimulation by light. V. The relation between intensity and critical frequency for different parts of the spectrum. Journal of General Physiology 19:965-979. MacLeod, D.I.A. 1972 Rods cancel cones in flicker. Nature (London) 235:173-174. 1974 Signals from Rods and Cones. Ph.D. dissertation. University of Cambridge. Sharpe, L.T., and K. Nordby 1984 Receptor function and interaction in a rod monochromat. Perception 13:A7-A8.
MOBILITY AND VISUAL PERFORMANCE UNDER DIM ILLUMINATION Ian L. Bailey This paper addresses two aspects of visual performance--namely visually guided mobility and reading. Both of these are dependent on vision and become impaired under dim illumination. Experience with the mobility problems of low-vision patients can perhaps give some insight and understanding of the visual difficulties that normally sighted individuals may have at mobility tasks when the illumination is low. Regarding reading skills, much is known about changes in visual acuity or resolution limits with illumination, but changes in efficiency of reading has been relatively ignored. In this paper some experiments will be described that might relate to the prediction of impairments of reading efficiency under dim illumination. In their book Foundations of Orientation and Mobility, Welsh and Blasch (1980) described mobility as "the ability to move independently, safely, and purposefully through the environment." To travel from one place to another, an individual must have some information about the relationship of the starting point, the destination, and the route or alternative routes between them. Vision may help in recognizing key landmarks, identifying appropriate courses, establishing and maintaining an awareness of the general structure and geography of the environment, and identifying the desired destination. Under conditions of dim illu- mination, persons with normal vision experience some reduction of their visually guided mobility skills. Generally, more concentrated atten- tion is required for the visual search and identification tasks. Typically, the key tasks do not involve the resolution of fine detail but rather the detection of large objects or features. The more important objects or features usually have sharp edges or junctions. In an outdoors urban environment, one gets one's bearings and selects one's course by being able to see the horizon, hills, buildings, cars, roadways, poles, trees, etc., seen in contrast silhouetted against brighter or darker backgrounds. In negotiating one~s immediate path- way, the pavement must often be distinguished from curbs or grass bor- ders. Obstacles or potential hazards may be identified if they or their shadows exhibit a contrast difference with the path or the sur- rounding area. Indoors, under dim illumination, critical cues for mobility come f rom the identif ication of junctions between walls, floors, and ceiling. Windows, doors, furniture, etc., create visual discontinuity against the walls, floors, or ceiling. For mobility 220
221 large objects or features must be detected, and it is contrast borders between objects and their immediate surroundings that enable their detection. The borders involved are typically sharp, and relatively seldom is it important to notice gradual changes in luminance across a large object. Examples of useful gradual changes in luminance might be the detection of undulations in snow, in sand, or on an uneven pave- ment. Contrast sensitivity, at least for the detection of borders, can be taken to be a critical attribute for visually guided mobility. Marron and Bailey (1982) studied a group of patients with low vision by examining relationships between their mobility skills and their vision characteristics. There were 19 subjects, all of whom were legally blind and enrolled in a residential rehabilitation program at the Western Blind Rehabilitation Center of the Veterans Administration Medical Center at Palo Alto, California. It should be emphasized that this was a select population group--it cannot be taken as typical of the low-vision population. These patients were in a rehabilitation program because they had been having difficulty coping with problems in their everyday lives. Visual acuity, contrast sensitivity, and visual fields of these 19 subjects were measured, and their mobility skills were evaluated on two different kinds of mobility courses. The results showed that, within this sample population, visual acuity had very little predictive power for mobility performance. Both visual field size and contrast sensitivity, however, did show moderate predictive power. Their predictive power was about equal, but a combination of the two gave a substantially improved prediction of mobility perfor- mance. Examination of the data f rom individual subjects in the Marron and Bailey study (1982) indicates that mobility performance consistently becomes impaired when the field is reduced to 20 degrees or less. Pelli and Serio (1984) , at Syracuse, have made measurements of mobility performance on normally sighted subjects who wore optical devices to create restrictions of their visual capabilities. He simulated concen- tric field defects by using spectacle-mounted tapering cones. Such sim- ulations of concentric field loss are somewhat deficient in that the restricted fields do not move with the eyes, but nevertheless, such devices can provide useful insights into the problems that result from visual field restriction. Pelli and Serio found that mobility started to become impaired when the visual field was reduced to a diameter of 7 degrees. This result is moderately compatible with those of Marron and Bailey (1980), especially considering that the partially sighted group of Marron and Bailey invariably had some other loss of function accom- panying the visual field loss. From these two studies, it seems reason- able to suggest that concentric field losses alone might be expected to produce mobility difficulties should the field become restricted to about 10 degrees diameter or less. Contrast sensitivity is the other critical feature of vision that Marron and Bailey identified as being important for mobility. A con- siderable amount of effort was spent trying to determine which feature of the contrast sensitivity function best correlated with mobility per- formance. They considered total area under the contrast sensitivity curve (CSF) curve, areas under selected portions of the curve, and
222 loom . - ,oO~ . _ In c °10 _ __ C_ O_ C)_ I 1 1 1 ace, V V\ aft\ V. o/ \\ : | ~ ~ I I I ~ ~ ~ ~ ~ ~ ~ ~1 ~I ! ~ ~ ~ ~ t; 0 ~5 t0 20 ~0 '~ Spatia I f r equency, c/deg FIGURE ~ Contrast sensitivity functions at f ive adaptation levels ranging from 0.0017 to 17.0 cd/m2. Source: De Valois et al. (1974~. slopes on both the high- and low-spatial-frequency sides, but it was the height of the peak of the CSF curve that showed the strongest statistical relationship to mobility. Studying their raw data, it appears that mobility problems can be expected when the contrast sensitivity becomes reduced to about 1.2 log units (or 24 dB). Pelli and Serio (1984) studied the effect of induced contrast sensitivity losses with normal subjects and found that their subjects exhibited mobility deficits when the degradation of vision reduced the key contrast borders in the environment to 0.8-1.5 log units (16-30 dB). Pelli and Serio's results agree moderately well with those of Marron and Bailey (19801. They also agree with my own clinical experience. For the past 12 years or so I have been making measurements of contrast sensitivity on many patients attending low-vision clinics. My impres- sion has been that persons begin to have significant mobility difficul- ties when photopic contrast sensitivity reduces to l.S log units (or 30 dB) or lower. Contrast sensitivities of about 0.5 log units (lO dB) or less usually prevent useful visually guided travel. Contrast sensitivity is undeniably relevant to visually guided mobility. To consider the mobility difficulties that normally sighted individuals suffer under reduced illumination, it may be useful to look at changes in contrast sensitivity as a function of f ield luminance. Figure l shows results f rom De Valois et al. (1974~. Figure l shows five contrast sensitivity functions measured under different screen luminance levels ranging from 0.0017 to 17.0 cd/~. The second upper- most curve here was taken at a level that H.W. Leibowitz (in this vol- ume) referred to as civil twilight, and the lowest curve represents conditions of nautical twilight, the level at which navigators have difficulty identifying the horizon. Under civil twilight conditions, where some limited reading is still possible, the peak is quite high, but as the nautical twilight is approached the peak contrast sensitivity
223 reduces substantially to a height of about 1.0 log units (or 20 dB 10 percent contrast). A person's visually guided mobility depends on the contrast of the more relevant contrast borders in the environment. However, there is enough circumstantial evidence to suggest that some mobility difficulty can be anticipated if border contrast sensitivity under the prevailing conditions is 20 dB or less, and at a contrast sensitivity of 10 dB or less visually guided mobility might be expected to be extremely limited. Contrast sensitivity functions are usually measured using sinusoidal gratings, but it is the peak of the curve that seems to be most critical in determining whether an individual has sufficient contrast sensitivity for mobility purposes. A functionally relevant test of contrast sensi- tivity might be to test contrast sensitivity with simple straight-edge contrast borders rather than with a series of sinusoidal gratings of differing spatial frequencies. I would like to draw attention to a rather ancient and overlooked test of contrast sensitivity that was devised by a British ophthalmologist, George Young (19187. The George Young Threshold Test is a small booklet containing 36 test plates, each being 10 x 15 cm. On each white plate there is one 15-mm diameter spot of ink, and from page to page the ink of the spot becomes progressively more diluted so that the contrast of the spots reduces systematically. mere are 8 pages with black or gray spots and 7 pages each for a red, green, blue, and yellow series of test spots. The subject's task is to locate the spot on each page. Young (1918) emphasized that this test was very sensitive for the early detection of glaucoma. Presented at a close distance, the George Young Threshold Test presents a test that is essentially similar to a test of border contrast sensitivity. I ac- suired a copy of this test in 1963 and I have periodically used it in the clinic since then. Influenced by the George Young Threshold Test, Tom Raasch and ~ have produced a contrast sensitivity test (Figure 2) that uses photographically produced strips, each 2.5 x 35 cm, along which there is a contrast border--a straight line junction between dark and light. Subjects are required to identify the location of the border on each strip. In all there are 25 strips, and there are S strips on each of 5 pages. m e range of contrast extends from 0.2 to 3.0 log units S4 to 60 dB), and over most of the range, the increments are in 0.1-log-unit (2 dB) steps. A second component of our test is a comparison of measurement of visual acuity at different contrasts (approximately 100 and 10 percent). The difference between the high- and low-contrast visual acuities indicates the extent to which visual acuity is degraded as the contrast is reduced. Theoretically, relating our test results to the contrast sensitivity function, the border con- trast sensitivity score should approximate the peak of the contrast sensitivity function and the difference between the high- and low-acuity scores should provide a measure of the slope of the CSF as it approaches the high-spatial-frequency cutoff. The high-contrast visual acuity score should be an index of the cutoff spatial frequency. We have made eight copies of our test but have been unsuccessful in economically reproducing the test to the required level of precision. This kind of border contrast sensitivity test is a more realistic representation of the contrast discrimination tasks encountered in everyday mobility, and
224 . . At.. ~,~i. it... · _~. _. ~ ~.1 * U. · ~% : ~ : ~ FIGURE 2 Bailey Contrast Sensitivity Test. me border contrast sensitivity section of the Bailey Contrast Sensitivity Test. it may be better than grating-based tests for the purpose of predicting functional performance. The test is very easy to use and has surface validity and apparent relevance to both the clinician and the patient. The test should be sensitive enough to identify all but the rarest abnormalities of contrast sensitivity that result from diseases to the visual system. Even though aging and night vision will be dealt with in greater detail in another paper (C. Owsley, this volume), I should emphasize that older people do have substantially more difficulty under dim light conditions. Older individuals have less light reaching their retina, partly because the ocular media absorbs more light and partly because the pupil is smaller. If one were to consider retinal illuminance alone, one could think of an old pair of eyes as being similar to a young pair of eyes wearing sunglasses when the conditions are photopic. But when the conditions are scotopic, it is as though the older eyes are wearing a double layer of sunglasses or, perhaps, welding filters. Figure 3 shows typical dark adaptation curves for a 40- and a 70-year- old person. The curves are of similar shape, but they show that a 70 year old takes much longer to adapt to a dimmer luminance level. For example, if one considers the level of sensitivity for detection at which the 40-year-old person will be operating after 11 min of dark adaptation, it will take the 70 year old 18 min to develop the same degree of sensitivity at that same luminance level. In any considera- tion of night vision, special attention must be paid to the strong effects of age.
225 cat J . _ 3 _ 2 Log Ii \ \ \ ;\ 1\ 1\ o I \~\\\,: in yrs 1 1 \J~ \ 70-82 I ~ i\ \ ~ i ; \~ 60-69 1 1 1 40-59 1 1 1 I'm ~ 1 1 10 20 30 Tlmc in minutes FIGURE 3 Graph showing why older people take longer to reach a given level of dark adaptation (after Birren and Shock, 1950~. Source: Peale, 1963. Under night vision conditions it should be expected that both reading acuity and reading efficiency will be impaired. To consider functional reading ability under dim light conditions, attention should be paid not only to the size of the smallest type size that is legible at all but also to the type size required for efficient reading per- formance. Here I will describe some experiments that I have conducted with Sam Berman and Robert Clear of the Lawrence Berkeley Laboratories. We looked at reading efficiency and its relationship to type size, luminance, and contrast. As our reading task, we used 40 Bailey-Lovie Word Reading Charts (Bailey and Lovie, 1980) (Figure 4~. Each of these charts contained samples of 17 type sizes. On the 11 lower rows, there are 6 words per row (two 4-letter, two 7-letter, and two 10-letter words). The print size ranged from 80 point (10 M) to 2 point (0.25 M). With the chart at 40 cm it is theoretically predicted that the reading of the smallest row would require a visual acuity equivalent to 20/12.5 (0.40/0.25 M). The size progression on these charts is loga- rithmic, with each row being about 0.8 times the size of the row above. The charts were prepared as clear transparencies and viewed against a retroilluminated opal screen. The luminance of the background screen could be varied according to the number of incandescent bulbs engaged and by the setting of a rheostat. An identical light box and opal screen were positioned above the subject's head so that with reflectors it could be optically superimposed on the light box with the test chart A disk containing an open window and two apertures with partially re- flecting mirrors was introduced close to the observer's eyes. The
226 ~zs~' ~ ~ answers plnx securities disease luck collection navy dynamic additional incredible briefly gate veteran encouraged lane is historians gold carries membership bullets edge me managed attempting stem fine remembered crawled 12 Stitch prooodu~ dew outdoor fail ~ FIGURE 4 Bailey-Lovie Word Reading Chart (Bailey and Lovie, 1980~. Illustration shows 1 of a series of 40. observer could view the test charts through the open window to achieve maximum contrast (Michaelson's ratio of 97 percent) or through the two partially silvered mirrors to achieve contrast levels of 64 and 17 per- cent. For the experiment described here, the subjects read silently while eye movements were monitored using an infrared reflection system. Reading speed was measured for up to 11 different type sizes over a luminance range that extended from 3 to 1,600 cd/m2. The measurements were made on 17 normally sighted young adult subjects. In analyzing the data, each subject's reading speed was normalized by first averaging the reading speed with large (12- and 16-point) type, as this could be read under every one of the various conditions tested. Then, for each subject all other speeds were expressed as a ratio to his or her average speed for the large type sizes. As expected, all subjects showed a slowing of their reading speed as threshold type size was approached. Under conditions of low contrast and low luminance, there was the expected reduction in the ability to resolve the small print. Figures 5a and 5b show how, on average, the reading efficiency for 10-point (magazine) and S-point (classified advertisement) type changed as a function of luminance for the three contrast levels tested. For the 10-point type maximum speed could be achieved with either the high or the medium contrasts over almost the entire luminance range. The reading efficiency for the low-contrast material was only marginally reduced, with the reduction being most pronounced when the luminance was low. For the 5-point type, the reduction in contrast had an enor- mous effect on reading efficiency. Even the medium contrast charts produced a substantially slower reading speed for all luminances. These results are not startling, but they do strongly illustrate that reading efficiency is very dependent on type size and that size and efficiency relationships can change dramatically when contrast or luminance is reduced.
227 ~ 2 e_ A) ·_ hi, 1 5 i_ ~ 1 a MEAINI RELATIVE READING TIME VERSUS LUMINANCE L = ~ ~ o/O c ontrast M = 64 °/0 contrast H = 97°/0 contrast for 10 point print 3 10 22 46 100 250 630 1600 Cd/ ~ b MEA N RELATIVE READING TIME VERSUS LUMINANCE ``Low for 5 point print M 10 22 46 100 250 LUMINANCE ( Cd/ M2 ) 630 1 600 FIGURE 5 Reading time versus luminance functions for three contrast levels and for two type sizes. Data represent averages of normalized reading times per row for 17 normally sighted young adu It subjects using a 40-cm viewing distance. ( a) Type size is 10 point ( 1.25 M) . (b) Type size is 5 point (0. 63 M) .
228 The most interesting result to come from this study emerged when each of the various reading efficiency versus type size curves was plotted in a manner such that the type size was expressed as a ratio to the threshold type size. This normalizes type size relative to thresh- old for each condition. In Figure 6 we present curves for low-contrast/ low-luminance conditions, medium-contrast/medium-luminance conditions, and high-contrast/high-luminance conditions. The three curves are vir- tually identical in both shape and location. The very small lateral displacement between the three curves is probably an artifact that can be attributed to an underestimation of threshold type size under the better viewing conditions because some subjects were able to read all words on the charts, and consequently, their resolution limit about which the curves were normalized was underestimated. Taking these curves to be identical leads to the conclusion that people can perform this particular reading task with maximum efficiency, provided that the type size is about 2.5 to 3 times larger than the threshold size. This is the case regardless of illumination and contrast. This suggests a useful rule for illuminating engineering. For illumination to be ace- quate for reading tasks, there should be at least a 2.S to 3 times reserve above the threshold of resolution. In practice, a good test of adequacy of illumination for a specific task might be to determine whether an individual can perform the required tasks at a distance that is 2.5 to 3 times the normal operational viewing distance. Printers have long been aware that some reserve over resolution threshold is required for efficient and acceptably comfortable reading performance. Economic considerations encourage printers to use smaller type sizes, but they are limited by the public's willingness to accept the smaller type. It has evolved that 8-point type or thereabouts is used for most newspaper material. This print is normally read at about 40 cm. The distance at which newsprint is just legible to normally sighted indi- viduals is just beyond 1 m. This fits in well with our findings that suggest the desirability of a reserve of 2.5 to 3 times above the threshold. In summary, visual performance at mobility and reading tasks under dim illumination have not been specifically addressed in the currently available literature. Two limited studies directed at understanding mobility skills of persons with low vision suggest that visually guided mobility skills are likely to be somewhat reduced if maximum contrast sensitivity is only in the 1~0- to 1.5-log-unit range. For contrast sensitivities below 0.5 log units, visually guided mobility skills can be expected to be minimal. It has been argued here that a test of border contrast sensitivity under the given illumination conditions might be a more relevant test of visual capacity for adequate mobility. To determine whether an individual is able to perform reading tasks in dim light, it is a simple practical matter to measure the threshold type size for legibility under the illumination conditions of interest. Based on photopic studies reported here, it might be expected that a reserve of three times the threshold type size might be necessary to achieve maximum efficiency under mesopic conditions.
229 , \ ', \\ '\\ \ \ ''\\ ~ \ . ~ 2 .E · _ hi;, 1 5 . _ - ~ 1 CHANGE IN READING TIME AS THRESHOLD SIZE IS APPROACHED ~ L ·17 10 At\\ . \ ·\ \ %\ \ "N \ a. 64 46 ·9 7 250 1 2 4 ~ ratio- print size relative to threshold FIGURE 6 Normalized reading times versus ratio of type size read to threshold type size. Data from 17 young adult subjects reading Bailey-Lovie Word Reading Charts under three conditions of illumination (L). REFERENCES Bailey, I.L., and J.E. Lovie 1980 The design and use of a new near vision chart . Amer ican Journal of Optometry and Physiological Optics 57: 378-3&7. Birren, J.E., and N.W. Shock 1950 Age changes in rate and level of dark adaptation. Journal of Applied Physiology 2: 407-411. De Valois, R.L., M. Morgan, and D.~. Snodderly 1974 Psychophysical studies of monkey vision. III. Spatial luminance contrast sensitivity tests of macaque and human observers. Vision Research 14: 75-81. Marron, J.A., and I.L. Bailey 1982 Visual factors in or ientation-mobility performance. Amer ican Journal of Optometry and Physiological Ontics 59: 413-426. Pelli D.G., and J.A. Serio 1984 The visual requirements for mobility. ARVO Abstract Investiga- tions in Ophthalmology and Visual Science 25(Suppl.) :72.
230 Weale R.A. 1963 The Aging Eye. London: Lewis. Welsh R .L ., and B .B . B lasch 1980 Foundations of Orientation and Mobility. New York Foundation for the Blind. : Amer ican Young, G. 1918 Threshold tests. British Journal of Ophthalmology 2 :384-392 ~ .
GENERAL DISCUSSION - ALEXANDER: I'd like to ask Dr. MacLeod if he'd care to speculate on the mechanism that might underlie this two-system rod response. It is in the photoreceptors, or is it further on downstream? MACLEOD: That's obviously an important question, and unfortunately I'm at a loss to give you any useful information as to the answer. We've done our best to distinguish the two mechanisms spectrally or in terms of the directional sensitivity, but we never find any distinction between them, There are several hypotheses that are attractive. The one that's most attractive, I think, is that there are fast cone path- ways to the ganglion cell and slow rod pathways. I've tried to test that by putting masking flickering stimuli in through the cones to try to sort of block the fast cone pathway and upset the null between the fast and slow rod mechanisms, but we haven't so far succeeded in up- setting the null. So I just don't know. There is some relevant receptor physiology. In the skate, D.G. Green and I. Siegel (Science 188:1120, 1975) found a two-branched CFF log intensity function, but so far that has not been reported in the rods of other animals. MAKOUS: I'd like to comment on Dr. Bedell's suggestion that the Stiles-Crawford effect might account for some of the unexpectedly low effects or small effects of defocus in photopic vision. One can apodize the pupil so that transmission is reduced toward the edges. Of course, if you defocus an image that passes through an aperture like the pupil of the eye, naturally you get reduced optical transfer, but the inter- esting thing is, an apodized pupil produces a uniformly higher transfer function for a defocused image. So what the Stiles-Crawford effect does is effectively apodize the human pupil and produce superior con- trast sensitivity for defocused images. This applies even to images that are defocused by a small amount corresponding to about a quarter of a diopter. PITTS: I appreciate Dr. Haber's paper very much, because I've published in this area some years ago while I was in the Air Force. But I'd like to suggest to him that in addition to losses of visual input (which he emphasized), there's also increased confusion and illusory effects as we get in dim illumination. Dim illumination is not the reason for this increase in confusion and illusory effects, because it is even worse under bad weather conditions. 231
232 Regarding some of the examples that he gave, I would also like to say that the stars are not fixed for a pilot at night, particularly for a fighter pilot flying alone. If a fighter pilot does not look at his instruments regularly, that star will move. We have many a pilot that's locked onto a star, and you have one hell of a job trying to convince him that it's not a light that he's looking at, that it's a star he's flying toward. The other problem is that as you decrease light at night, the patterns that we think are set up with the lighting of the systems can be confused. There is a particular time when pilots con- fuse lights in the sky, called stars, and lights on the ground, that are streetlights. They have even landed aircraft low; Japan Airlines 2 or 3 years ago landed in San Francisco Bay just because of these sorts of problems. And so they are real; they're there; and as you lower the light levels, they get worse. As you decrease the weather conditions, they get even worse. HABER: I didn't touch on illusory perceptions at all, but those are obviously very relevant. Your last example shows that as you reduce illumination, and particularly when flying over boundaries between water, cities, and sky, the pilot has to deal with three horizons, the clear ones being the wrong ones to fly to. He is likely to confuse the clear ones with the real one and impact the ground when he didn't expect to. In my comment about the visible stars not moving, I was being an astrophysicist there rather than a perception psychologist. The stars clearly move in that environment if you try to fixate or track them with your eyes--the familiar autokinetic effect. Clearly I should add those to the kind of presentation I gave. WATSON: If I read your graphs correctly, Don [MacLeod], you only showed that a reduction of a factor of about 2 from the peak temporal sensitivity to the lowest chopper is going to reduce the amplitude of your signal by a factor of 2. So it seems to me you'd be back to where you started out. Indeed, the curve may go down much more than you showed it at the low end. The second question has to do with your description of phase lags and cancellation techniques. You treated everything as though it were linear, and I wanted to know if you'd thought through the consequences of very early nonlinearities and how that would affect the ability to cancel things travelling through the two pathways? MACLEOD: On the first point--that there's not a sufficiently severe loss of sensitivity at low temporal frequencies to make use of the chopper worthwhile--it's true that at least some of the curves I showed didn't have that much of a loss at low temporal frequencies. But I omitted to tell you--and will apologetically tell you now--that those were actually square-wave temporal modulations. That makes a difference. For sine waves, you can get a factor of 10 between 1 Hz and 10 Hz and in that case, for a sufficiently low spatial frequency content, I think, you might expect a fivefold improvement (10 divided by 2) from using the chopper. And with an image converter that inverts contrast instead of chopping, the factor of 2 would not have to be lost. There is a problem that with good edges in the stimulus, you have-- by virtue of fixation of eye movements--some sensitivity down to zero
233 temporal frequencies, without any need for chopping. But when the edges of the object are obscured by fog or haze or perhaps even retinal resolution losses, then it might be helpful. I expect that the domain of usefulness might be only below about 1 cycle/degree for rods, and a bit above that for cones. On the second point--about the role of nonlinearity in the pulling of rod and cone signals--I did show a slide where a model that included the assumption of linearity seemed to fit the data fairly well, although not perfectly. So I think that shows that the consequences of nonline- arity of the signals are not tremendously large. In principle, though, you would expect that second ha``,~onic components in the rod and cone signals would not cancel along with fundamental components. That would lead to residual flicker in the conditions of the mesopic null. I think there are some observations which do show that there is this sort of distortion--one of them being that under some conditions there are two phases that generate minimal flicker as you vary the phase between a rod stimulus and a cone stimulus. JOHNSON: I'd like to have Dr. Haber clarify a point with regard to night visual approaches. The work that Conrad graft has done with night visual approaches indicates that there are certain situations in which pilots do not check their instruments and will come in too low on their final approach because they think that they are higher than they actually are. The common components of these situations include a body of water or other dark expanse in front of the airport with a well-lit city in the background. This creates a misperception that causes the pilots to believe that they're actually higher than they are. What you described, in terms of the problems with training and going from day- light visual approaches to nighttime visual approaches, is exactly the opposite. Pilots flare too high, the presumption there being that the Pilots think they are actually lower than they are. I was wondering if there was some kind of explanation that could account for these seem- ingly opposite effects? HABER: Yes, the two tasks are quite different. In the approach over water--which were the examples by Conrad Kraft, who is the one who has done the most work with the Boeing simulators on that--you have the basic problem of identifying which horizon is the true horizon. And if the pilot judges the horizon to be the water-land horizon, then he's going to be aiming below that, because obviously the runway must be below that horizon. So he comes in much too low. That' s not the only characteristic that accounts for that kind of accident. The general characteristic for night f lying, particularly for beginning pilots--where you don't have a horizon at the near end of the runway but where you've got city lights all around you--is simply having trouble judging how high you are. So they will come in too high over the numbers at the end of the runway, where they should be ready to flare and touch. Instead, they may be a few feet too high (5, 10, or even 50 feet) depending on the type of plane and type of approach, and then they float down the runway. That's obviously a better outcome than landing short, as long as the runway's long enough and they don' t bounce too hard on it. Thus, the two tasks have quite different courses and differ ent ou tcomes.
234 M. O'NEAL: I'm at the Aerospace Medical Research Laboratory, at Wright-Patterson AFB, Ohio. This is directed to Dr. Harvey. Contrast sensitivity has been touted recently as a replacement for the acuity chart. Recently, at a meeting at Wright-Patterson, it was suggested that contract sensitivity testing could be used to select pilots for specific missions. In your opinion, do you feel that contrast sensi- tivity testing has progressed to the point where this is feasible and proper? Also, what research is going on in contrast sensitivity per- ception to correlate it with performance? HARVEY: Let me say that the so-called contrast sensitivity func- tion is only predicting a small number of operational things and is not exactly correct. I mean, if you look in the vision literature, maybe 50 percent of the articles in vision research are devoted to some aspects of the relationship between spatial frequency characterization of stimuli and some sort of functional properties of the stimuli--some sort of performance in terms of discrimination, detection, and whatnot. So there's an incredible literature that relates contrast sensitivity to basic psychophysical functions. In terms of actual real-world tasks, I don't know right now what's going on in the Air Force. Are there actual attempts to set up a proper evaluation of these contrast sensitivity functions in relation to tasking? There was a lot of evidence reviewed recently by the Committee on Vision--that I've described--suggesting that the contrast sensitivity function could become a better tool for selection if used in the proper way. We recommended that research be carried out comparing individual differences, not only on contrast sensitivity but on a number of basic measures to look at the relationship between individual differences on these tasks and individual differences on the performance of some tasks which are of interest to the Air Force.
