Flash Blindness Recovery of a Tracking Task on Cockpit Attitude Indicators
INTRODUCTION: Intense light exposures can cause temporary flash blindness, degrading pilot performance during flight. The present study investigated factors influencing time to recover from flash blindness for tasks resembling aircraft control using an attitude indicator. Prior similar studies of flash blindness used only reflective gauges whereas modern cockpits include emissive displays, so recovery differences between reflective and emissive instrument types were of interest as was the influence of varying ambient luminance levels.
METHODS: Nine subjects performed attitude indicator horizon stabilization and tracking tasks on both a reflective and an emissive attitude indicator. Subjects were exposed to short (150 ms) high intensity broadband light flashes at three retinal exposure levels [6.5, 7.0, and 7.5 log troland-seconds (logTd·s)] beforehand. Additionally, ambient luminance was manipulated (1 cd · m−2, 10 cd · m−2, and 100 cd · m−2). The time to level the horizon after a flash exposure was measured. After leveling, roll and pitch errors made while maintaining straight and level flight by countering added perturbation were also tracked.
RESULTS: Greater flash intensity usually increased recovery time. For the reflective attitude indicator, as ambient luminance increased, flash intensity had weaker influence on recovery times, with recovery times ranging from 6–30 s. For the emissive attitude indicator, however, ambient luminance did not appreciably influence recovery times, with recovery times ranging from 8–16 s.
DISCUSSION: The reflective attitude indicator was more advantageous for flash blindness recovery in high (100 cd · m−2) ambient luminance and the emissive indicator was relatively more advantageous in low (1 cd · m−2) ambient luminance.
Arizpe JM, McAnally RE, Hart MV, Kuyk TK, Smith PA, Goettl BP. Flash blindness recovery of a tracking task on cockpit attitude indicators. Aerosp Med Hum Perform. 2025; 96(12):1032–1042.
Exposure to intense broadband (white) light sources may result in flash blindness, a temporary reduction in visual sensitivity. The light sources may be viewed directly or indirectly from reflections off clouds, water, windows, or solar panel arrays, and the resultant temporary visual impairment can last seconds or several minutes. Lasers are another intense light source that could cause flash blindness. The temporary loss of vision is caused by saturation of the photoreceptors and visual afterimages. While temporary, the loss of sensitivity in central vision can prevent pilots from accurately reading flight instruments used to maintain control of their aircraft.
The U.S. Air Force has a history of studying the effects of intense light exposures on human vision. Previous research has identified source intensity,1 retinal size,2–4 and ambient luminance1 as key factors influencing flash blindness. For example, recovery times increase with flash intensity and flash size, but they decrease in increasing ambient light. Work on flash blindness led to a model that provides a visualization of how an image appears during the recovery process.5–7 Additionally, lasers have been used to cause flash blindness with recovery times consistent with the model.8–10
Another important factor is whether a cockpit instrument is reflective or emissive. Early studies used reflective gauges found in cockpits of the time. To be seen and interpreted, reflective gauges require an external light source (e.g., sun or cockpit lighting) to illuminate and reflect off painted dials, needles, and printed letters. Aircraft gauge technology has evolved and modern cockpits have both reflective and emissive flight instruments. Emissive instruments do not rely on reflected illumination to be visible. Although recent studies have used emissive displays in flash blindness recovery experiments only one side-by-side comparison with reflective and emissive instruments has been done. Arizpe et al.11 showed ambient light significantly impacted Landolt C orientation detection recovery with reflective instruments, but not with emissive instruments. We hypothesized similar results would be obtained on two tasks involving interpreting the tilt of the horizon (roll) on an aircraft attitude indicator and controlling pitch and roll to maintain level flight. These tasks differ from the earlier task as they do not test the ability to resolve fine detail. Instead, the first task tests how long after a flash it takes an observer to determine if a large feature, the chromatic boundary between the ground and air that defines the horizon, is tilted (oriented) right or left and recover from the roll to level flight using a joystick. The second task tests if a bright flash has a long-lasting effect on flight performance by measuring the ability to maintain level flight when perturbations are introduced that require corrective action.
We hypothesized that attitude recovery time with emissive instruments will be affected less by changes in ambient luminance because they emit light, whereas the ability to read reflective flight instruments will vary depending on the illumination.12,13 Thus, we predicted that as ambient luminance fell, recovery times: 1) would increase for reflective instruments because the horizon on the attitude indicator becomes darker and more difficult to see; and 2) would remain constant for emissive instruments because the letters remain at a more constant brightness. We also expected flash blindness recovery times would increase with flash intensity. These hypotheses are consistent with those in Arizpe et al. for a static spatial resolution task and the same broadband flash and ambient light conditions as the present study.
Modern cockpits primarily use emissive displays, with gauges as back-up instruments. It is crucial for pilots to know which instrument type becomes visible faster after a flash and under what conditions.
METHODS
Subjects
A total of nine subjects (eight men, one woman) participated in this study. None had any flight experience. Mean subject age was 37.7 yr (SD 10.7). The study protocol was approved in advance by the Air Force Research Laboratory Institutional Review Board (Protocol Number: FWR20200191H). Each subject provided informed consent before participating.
Procedure
Flashes were presented in Maxwellian view to subjects’ right eye. A white light emitting diode (LED; MCWHL6 6500K, Thorlabs Inc., Newton, NJ, United States) was filtered to reduce blue light emission (>5 optical density for 270–351.2 nm and >6 optical density for 351.2–457.9 nm) and collimated using three lenses, including a telescope formed with two 70-mm focal length lenses. To focus the light to a 4-mm spot at the pupil, the light passed through a 17-mm aperture and a 100-mm focal length lens. Within the 100-mm focal length, a beam splitter reflected 70% of the light and directed it 90° to the pupil. Subjects viewed flight instruments through the beam splitter, which transmitted 30% of the light from a reflective attitude indicator and 40% of the light from an emissive attitude indicator (the transmittance difference was due to the polarization of the beam splitter and emissive display light). Unless stated otherwise, all luminance and flash intensities were measured at the plane of the subject’s pupil (i.e., reflecting the effect of the beamsplitter).
Subjects’ heads were stabilized using a head rest and bite bar. Pupil position was monitored with a camera to ensure the focused beam entered the subject’s pupil. Prior to a flash exposure the subject’s pupil was aligned with a circle on the camera feed that indicated the Maxwellian focal point, and images captured during flash delivery verified accurate alignment. Three different integrated retinal illuminances of the flash were used: 6.5, 7.0, and 7.5 logTd·s. Flash duration was 150 ms for all exposures and varying the LED brightness realized the different flash intensities, with the output spectrum verified as invariant. Flash illuminances were calibrated using the method described by Nygaard & Frumkes.14
Reflective and emissive instruments resembling modern attitude indicators were used as stimuli. The reflective attitude indicator was a gauge purchased from a hobbyist flight simulator company (Simkits, TRC Simulators b.v., Arkel, The Netherlands) and had blue representing sky, brown representing ground, and a white line highlighting the horizon (Fig. 1, left). When the aircraft rolls, the horizon tilts (left or right) to represent the movement, while pitch is represented by an insert behind the main face that translates up or down. A stationary, black reticle with orange highlights located in the gauge foreground serves as a reference for straight-and-level flight.
Citation: Aerospace Medicine and Human Performance 96, 12; 10.3357/AMHP.6563.2025
The emissive attitude indicator was a reproduction of the reflective attitude gauge colors, accents, and motion using the Unity game engine (Unity Technologies, San Francisco, CA, United States) and a customizable attitude indicator asset from the Unity asset store. The emissive attitude indicator differed slightly from the reflective indicator in that the entire face translated with changing pitch (Fig. 1, right). The emissive attitude indicator was displayed on a liquid crystal display monitor (ColorEdge CG319X, EIZO Inc., Cypress, CA, United States), with the backlight of the screen set to 40 cd · m−2 (i.e., 40 cd · m−2 is what full white would measure on the screen without the beam splitter).
A calibrated imaging colorimeter (LumiCam, Instrument Systems GmbH, Munich, Germany) was used to characterize the reflective attitude indicator, and the colors of the elements of the emissive indicator (e.g., white lines, blue “sky”, brown “ground”, orange miniature aircraft, etc.) were set to corresponding RGB values. Luminance ratios between elements were matched between emissive and reflective indicators to ensure comparable chromatic and luminance contrast. These ratios remained relatively constant despite changes in ambient luminance, although absolute luminance values did vary (see Table I).
The subject sat in a darkened room 57 cm from the reflective gauge and 76 cm from the emissive display. The distances were different due to configuration constraints, but the visual angles that each instrument occupied were equal from the subject’s perspective. The instruments were illuminated by two adjustable LED-panel light towers (LEDP260C, Godox Photo Equipment Co., Ltd., Shenzhen, China) flanking the subject position, casting light forward. Three ambient luminance conditions (100, 10, or 1 cd · m−2) were created by adjusting the light towers to illuminate a white card at the display plane to the same values. Fig. 1 shows the subject’s view of the attitude displays and equipment configurations. The luminance of regions surrounding the gauges ranged from near zero to approximately 26 cd · m−2 (excepting any glints or specular highlights) for the 100 cd · m−2 reflective display condition and scaled proportionately for the other conditions.
Trials had two sections. First, the subject viewed the attitude indicator in a stationary and level position (both pitch and roll). When the subject’s pupil was aligned, the experimenter triggered the flash. Coincident with the flash, the attitude indicator roll moved either 20° clockwise or 20° counterclockwise, with no pitch change. Because the flash obscured this movement, subjects could not anticipate the new gauge position. When the subject recovered sufficiently to identify the orientation of the horizon, they pressed the trigger button and immediately tried to level the attitude indicator using a flight joystick. Recovery time was measured as the time between the flash onset and when the subject pressed the trigger button. The time until the horizon was successfully leveled was also recorded as verification that the trigger button press corresponded to visual recovery.
The second section began once the subject had leveled the indicator’s horizon, whereupon a forcing function was introduced, causing pitch and roll changes. The amplitudes and frequencies of the forcing function were produced from a sum of sinusoidal functions to mimic the dynamics of an in-flight aircraft affected by factors like turbulence and changes in wind speed and direction. Subjects tried to keep the indicator level in both pitch and roll using the joystick for 30 s, after which the trial ended. Prior to data collection, subjects were allowed as many practice trials as desired (no flash) to become comfortable with the joystick sensitivity and attitude indicator response.
Each subject was exposed to 18 flashes in a 2 instrument types (Emissive, Reflective) × 3 flash intensities (6.5, 7.0, 7.5 logTd·s) × 3 ambient luminance levels (100, 10, 1 cd · m−2) design. Data were collected in two sessions, one for reflective and one for emissive instruments. Baseline data with no flash were also collected for each instrument type under each ambient luminance. After every flash trial, subjects rested in the dark for 5 min to allow vision to recover before proceeding. A pseudorandomized order of ambient luminance and flash intensities was implemented to mitigate order confounds.
Statistical Analysis
The recovery time data were first analyzed in an omnibus three-way repeated measures ANOVA with Instrument Type, Flash Intensity, and Ambient Luminance as within-subjects factors. Two-way repeated measures ANOVAs with Flash Intensity and Ambient Luminance as within-subjects factors and paired difference tests were used to characterize the significant interactions from the omnibus ANOVA. These two-way ANOVAs were performed separately for each display. Greenhouse-Geisser correction was performed when sphericity was violated for a given factor or interaction in an ANOVA. Effect sizes for ANOVA factors and interactions were reported as partial eta squared (ηp2). Compared to eta squared (η2), the effect size estimated for a given factor by partial eta squared is much less affected by the other factors in an experimental design, which makes partial eta squared an effect size estimate suitable for comparison across different experiments.
Tracking error was analyzed separately for reflective and emissive indicators due to differing pitch and roll step resolutions, precluding comparisons of absolute error differences. ANOVAs were conducted separately for roll and pitch error (absolute deviation in degrees) using two-way repeated measures with Flash Intensity and Ambient Luminance as within-subjects factors. Significant ANOVA effects were further characterized with paired difference tests.
All paired difference tests were within-subject and a Shapiro-Wilk signed rank test determined if differences were normally distributed. Because some difference distributions deviated from normality, for consistency, we performed the Wilcoxon signed rank test for all paired difference tests throughout. Effect sizes for paired differences were therefore reported as the matched rank biserial correlation (ρmrb). When we had an a priori hypothesis about the direction of differences in paired comparisons, we conducted one-tailed tests, otherwise we conducted two-tailed tests. According to convention, corrections for multiple comparisons were not performed for planned tests of our a priori hypotheses (i.e., lower ambient luminance increases recovery time and tracking error for the reflective attitude indicator, higher flash intensity increases recovery time and tracking error in general, and tracking error after a flash is greater than in the absence of a flash). For concision, we report only those statistical tests that directly relate to our specific research questions and hypotheses.
RESULTS
For the reflective attitude indicator, lower ambient luminance increased flash blindness recovery time for horizon orientation discrimination overall and increased the slope of recovery time vs. flash intensity. That is, increasing the intensity of the flash prolonged recovery time to a greater extent at a lower ambient luminance level than a higher ambient luminance level. Ambient luminance did not affect flash blindness recovery time for the emissive attitude indicator. Greater flash intensity increased recovery time for both the reflective and emissive attitude indicator. These patterns are apparent in Fig. 2 and demonstrated in the statistical analyses that follow.
Citation: Aerospace Medicine and Human Performance 96, 12; 10.3357/AMHP.6563.2025
The omnibus three-way ANOVA on recovery times for attitude indicator horizon orientation discrimination yielded a significant Instrument Type × Flash Intensity × Ambient Luminance interaction (Table II, upper block). This interaction indicated the way ambient luminance modulated the effect of flash intensity on flash blindness recovery time differed between the emissive and reflective attitude indicators.
Separate two-way ANOVAs for each instrument type characterized the difference between instrument types. The ANOVA for the reflective attitude indicator (Table II, middle block) yielded a significant Ambient Luminance × Flash Intensity interaction while the ANOVA for the emissive attitude indicator (Table II, lower block) did not, indicating that ambient luminance modulated the effect of flash intensity on recovery time for the reflective, but not the emissive attitude indicator.
The significant interaction between ambient luminance and flash intensity for the reflective attitude indicator appeared to be because lower ambient luminance caused an increase in the slope of the positive linear relationship between flash intensity and recovery time (see Fig. 2, left panel). To test for these differences statistically, we fit a separate regression line of recovery time vs. flash intensity at each ambient luminance level for each subject. On the slope values of these fit regression lines, we then performed Wilcoxon signed rank tests as paired (i.e., within-subjects) comparisons between adjacent ambient luminance levels (Table III, upper block). These comparisons yielded statistically significant differences indicating that the recovery time vs. flash intensity slope was greater at 1 cd · m−2 than at 10 cd · m−2 and greater at 10 cd · m−2 than at 100 cd · m−2.
Lower ambient luminance not only increased the slope of recovery time vs. flash intensity for the reflective attitude indicator, but it also increased flash blindness recovery time overall. Ambient luminance did not, however, change recovery time for the emissive attitude indicator. The main effect of Ambient Luminance was significant in the ANOVA for the reflective attitude indicator (Table II, middle block), but not in the ANOVA for the emissive attitude indicator (Table II, lower block). Planned comparisons on the marginal means of ambient luminance (i.e., averaged across flash intensity) for the reflective attitude indicator yielded significant differences in the hypothesized direction of lower ambient luminance causing longer flash blindness recovery time. Wilcoxon signed rank test results listed in Table III (second block) indicated that for the reflective attitude indicator, flash blindness recovery time was longer at 1 cd · m−2 than at 10 cd · m−2 and longer at 10 cd · m−2 than at 100 cd · m−2.
Flash blindness recovery time for attitude indicator horizon orientation discrimination almost always increased with greater flash intensity. The ANOVAs for both the reflective and emissive attitude indicators yielded a significant main effect of Flash Intensity (Table II). For the reflective attitude indicator, planned comparisons between adjacent flash intensity levels at each ambient luminance level allowed for investigating the possibility of interaction with ambient luminance. The results are listed in Table III (third block) and show the Wilcoxon signed rank tests indicated that recovery times were significantly longer with greater flash intensities, as hypothesized, except for the 7.5 vs. 7.0 logTd·s flash intensity in 100 cd · m−2 ambient luminance. For consistency, the same planned comparisons were performed for the emissive attitude indicator and are also listed in Table III (third block). Without exception, all comparisons yielded significantly longer recovery with greater flash intensities, as hypothesized.
Recovery times were longer overall for the reflective display than the emissive display under 1 and 10 cd · m−2 ambient luminance, but the opposite was true under 100 cd · m−2. Planned comparisons of recovery times between reflective and emissive displays on marginal means for ambient luminance (i.e., averaged across flash intensity) yielded significant differences at each ambient luminance level (Table III, fourth block); however, the sign of the mean difference dropped and crossed from positive to negative as ambient luminance increased, with the mean differences being 10.40 s, 1.87 s, and −5.36 s at 1, 10, and 100 cd · m−2, respectively. Thus, the highest ambient luminance reversed the relative recovery time trends between the reflective and emissive instrument types.
Flash intensity affected tracking during the straight-and-level task for both instrument types, while ambient luminance affected tracking only on the reflective attitude indicator, and there were no interaction effects between flash intensity and ambient luminance in either type of tracking error for either instrument type. These effects are apparent in Fig. 3 and are detailed in the results of the statistical analyses that follow. Additional effects revealed by the statistical analysis are that average roll error (i.e., absolute deviation from a straight indicator), but not pitch error (i.e., absolute deviation from zero pitch) increased with flash intensity for the emissive attitude indicator. For the reflective attitude indicator, both roll and pitch error increased only from the 6.5 to the 7.0 logTd·s flash intensity. Ambient luminance did not affect either roll or pitch error for the emissive attitude indicator, but both types of tracking error increased when ambient luminance decreased from 10 to 1 cd · m−2 for the reflective attitude indicator.
Citation: Aerospace Medicine and Human Performance 96, 12; 10.3357/AMHP.6563.2025
Results of the ANOVA for roll error during the tracking task for the emissive attitude indicator are listed in the lower block in Table IV and yielded a significant main effect of Flash Intensity but yielded neither a main effect nor interaction involving Ambient Luminance (both P > 0.206). The corresponding ANOVA results for roll error with the reflective attitude indicator (Table IV, upper block) yielded significant main effects of both Flash Intensity and Ambient Luminance, but not a Flash Intensity × Ambient Luminance interaction.
An a priori hypothesis was that greater flash intensity would increase tracking error. For roll error, the Wilcoxon signed rank paired comparisons on the marginal means of Flash Intensity yielded significant differences in the hypothesized direction between adjacent flash intensity levels for the emissive attitude indicator and between the 6.5 and 7.0 logTd·s flash for the reflective attitude indicator (see Table V, roll error metric, flash intensity blocks, for both reflective and emissive). We had also hypothesized that lower ambient luminance would increase tracking error for the reflective attitude indicator. The Wilcoxon signed rank paired comparisons on the marginal means of Ambient Luminance for the reflective attitude indicator yielded a significant difference in the hypothesized direction between 1 cd · m−2 and 10 cd · m−2, but not between 10 cd · m−2 and 100 cd · m−2 (Table V, upper block, roll error metric).
The same analytic approach was applied to the tracking task pitch error and the same a priori hypotheses determined the direction of the paired comparison tests. The ANOVA on pitch error for the emissive attitude indicator (see Table IV, lower block) yielded neither significant main effects nor an interaction, whereas the ANOVA for the reflective attitude indicator (Table IV, upper block) yielded significant main effects of both Flash Intensity and Ambient Luminance. The Flash Intensity × Ambient Luminance interaction was not significant, but marginal for the reflective attitude indicator. As with roll error, the Wilcoxon signed rank paired comparisons on the marginal means of Flash Intensity for the reflective attitude indicator pitch errors yielded significant differences in the hypothesized direction between the 6.5 and 7.0 logTd·s flash, but not between the 7.0 and 7.5 logTd·s flash intensities (Table V, upper block, pitch error metric). Again, as with roll error, the Wilcoxon signed rank paired comparisons on the marginal means of Ambient Luminance for the reflective attitude indicator pitch errors yielded a significant difference in the hypothesized direction between 1 cd · m−2 and 10 cd · m−2, but not between 10 cd · m−2 and 100 cd · m−2 (Table V, upper block, pitch error metric).
To interpret how meaningful the flash and ambient luminance effects just reported were relative to the baseline (i.e., no flash) data, we also conducted analyses involving the baseline data. The following three patterns can be seen in Fig. 4. 1) In the absence of a preceding flash, ambient luminance affected neither roll nor pitch error for either instrument type (black symbols and lines, and see Table IV upper and lower blocks), indicating that the increased roll and pitch errors under 1 cd · m−2 vs. the higher ambient luminance levels for all reflective attitude indicator flash conditions were due to the mutual effect of flash and low ambient luminance, rather than the low ambient luminance alone. 2) For the reflective attitude indicator (Fig. 4 left panels), all flash conditions at 1 cd · m−2 had greater roll and pitch errors compared to baseline, whereas the flash conditions tended not to differ significantly from baseline at 10 and 100 cd · m−2 (Reflective, 10 cd · m−2, 7.5 logTd·s was the only exception). 3) For the emissive attitude indicator (Fig. 4, right panels), the 7.5 logTd·s flash tended to cause roll and pitch errors that were higher or marginally higher than baseline, regardless of the ambient luminance level (pitch error for 100 cd · m−2 was the one exception). Emissive attitude indicator pitch error for the 7.0 logTd·s flash was also higher than baseline for both the 1 and 100 cd · m−2 conditions.
Citation: Aerospace Medicine and Human Performance 96, 12; 10.3357/AMHP.6563.2025
Each of these three patterns apparent in Fig. 4 are also reflected in corresponding statistical analyses. Ambient luminance did not have a significant effect on no-flash baseline tracking error. Separate one-way ANOVAs on baseline tracking errors with Ambient Luminance as the within-subjects factor were conducted for each instrument type (i.e., emissive and reflective) and type of tracking error (i.e., roll and pitch error). The ANOVA results for the roll and pitch error are in Table IV for the reflective and emissive attitude indicator in the upper and lower blocks, respectively. None of the ANOVAs yielded significant effects of Ambient Luminance (all four F < 1.956, P > 0.173) on baseline tracking error, indicating that the ambient luminance factor on its own did not modulate tracking errors. Thus, the increased tracking errors at 1 cd · m−2 vs. the higher ambient luminance levels for all reflective attitude indicator flash conditions (reported above) were due to the combined effect of flash and low ambient luminance.
For the reflective attitude indicator, flash conditions had significantly greater roll and pitch errors compared to the no-flash baseline at 1 cd · m−2, but did not tend to at 10 and 100 cd · m−2. Wilcoxon signed rank paired comparisons of the tracking errors of each flash intensity vs. baseline were conducted for each ambient luminance level, display type, and type of tracking error. With the reflective display at 1 cd · m−2, all flash conditions had roll errors (Fig. 4, top left) that were significantly greater than baseline (all z >2.073, P < 0.020, one-tailed, uncorrected, ρmrb> 0.778). The same was true for the pitch errors (Fig. 4, bottom left) of all flash conditions with the reflective display at 1 cd · m−2 (all z >2.429, P < 0.006, one-tailed, uncorrected, ρmrb> 0.911). For the 10 and 100 cd · m−2 conditions, however, neither roll nor pitch errors were greater than baseline for any flash conditions with the reflective display (all |z| < 1.362, P > 0.102, one-tailed, uncorrected, ρmrb< 0.511), with the sole exception of the roll error for the 10 cd · m−2, 7.5 logTd·s flash intensity condition (z = 2.666, P = 0.002, one-tailed, uncorrected, ρmrb= 1.000). Unlike the reflective display and consistent with previous analyses above, the corresponding collection of paired comparisons with baseline for the emissive display did not suggest a clear interaction effect between ambient luminance and flash intensity.
For the emissive attitude indicator, the 7.5 logTd·s flash tended to cause significantly or marginally greater roll and pitch errors than baseline (except for pitch error at 100 cd · m−2), and the 7.0 logTd·s flash caused significantly higher pitch errors than baseline for the 1 and 100 cd · m−2 conditions. Specifically, emissive display roll error (Fig. 4, top right) for the 7.5 logTd·s flash condition was significantly higher than baseline for the 10 and 100 cd · m−2 conditions (both z >1.835, P < 0.038, one-tailed, uncorrected, ρmrb> 0.688) and was marginally higher for the 1 cd · m−2 condition (z = 1.481, P = 0.082, one-tailed, uncorrected, ρmrb= 0.556). Pitch error with the emissive display (Fig. 4, bottom right) for the 7.5 logTd·s flash condition was significantly higher than baseline for 1 cd · m−2 (z = 2.192, P = 0.014, one-tailed, uncorrected, ρmrb= 0.822) and was marginally higher for 10 cd · m−2 (z = 1.599, P = 0.064, one-tailed, uncorrected, ρmrb= 0.600), but was not significantly different from baseline for the 100 cd · m−2 condition (z = 1.362, P = 0.102, one-tailed, uncorrected, ρmrb= 0.511). Emissive display roll errors for the 6.5 and 7.0 logTd·s flash conditions were not significantly different from baseline under any ambient luminance level (all |z| < 1.008, P > 0.102, one-tailed, uncorrected, ρmrb< 0.512). Pitch error for the 7.0 logTd·s flash condition with the emissive display was significantly higher than baseline for the 1 cd · m−2 and 100 cd · m−2 conditions (both z >1.835, P < 0.037, one-tailed, uncorrected, ρmrb> 0.689), but was not significantly different from baseline for the 10 cd · m−2 condition (z = −0.059, P = 0.545, one-tailed, uncorrected, ρmrb= 0.022). Finally, pitch error for the 6.5 logTd·s flash condition with the emissive display was not significantly different from baseline for any of the ambient luminance conditions (all |z| < 1.007, P > 0.180, one-tailed, uncorrected, ρmrb< 0.378).
DISCUSSION
The present study investigated how various factors influenced flash blindness recovery times in an attitude discrimination and recovery task, and an attitude maintenance task. Differences in flash blindness recovery for reflective vs. emissive flight instruments were of particular interest as a prior study found changes in ambient luminance modulated performance on a spatial resolution task when the flight instrument was reflective but had a lesser effect when it was emissive.11 The study concluded that a reflective flight instrument was advantageous for flash blindness recovery in high (100 cd · m−2) ambient luminance and the emissive instrument was advantageous in low (1 cd · m−2) ambient luminance.
Mean flash blindness recovery times for the attitude discrimination task ranged from 6–30 s for the reflective attitude indicator and 8–16 s for the emissive attitude indicator and were dependent on flash intensity and ambient luminance level. Consistent with our previous study, recovery times increased with greater flash intensity for both reflective and emissive attitude indicators and ambient luminance level modulated recovery time curves for the reflective attitude indicator such that as ambient luminance increased, flash intensity had a weaker influence on recovery times and the times were shorter overall.
At the highest ambient luminance level, recovery times for the reflective attitude indicator were significantly shorter than with the emissive attitude indicator. In contrast, for the emissive attitude indicator, ambient luminance level did not appreciably modulate recovery time curves, which is why the highest ambient luminance reversed the relative recovery time trends between the reflective and emissive instrument types. Specifically, higher ambient luminance shifted reflective attitude indicator recovery time curves downward and less steeply such that recovery time curves that were higher for the reflective than for the emissive attitude indicator at 1 cd · m−2 and 10 cd · m−2 were lower at 100 cd · m−2. Thus, neither indicator type consistently outperformed the other; the relative advantage depended on ambient luminance. The reflective indicator was more advantageous for flash blindness recovery in high ambient luminance, and the emissive indicator was more advantageous in low ambient luminance.
Standard safety practice for pilots is to dim cockpit lights during night takeoff and landing to allow them to see runway lights and other aircraft, reduce glare, prepare them for emergencies, and minimize eye strain. Military pilots may dim flight instruments during stealth, low-altitude, or reconnaissance flights to reduce the likelihood of being detected or tracked from the ground. Increasing the luminance of flight instruments could be a countermeasure to flash blindness, but a balance must be struck between cockpit instrument readability and maintaining a dim cockpit for safety or tactical reasons. Our data informs the trade-off between cockpit instrument readability and maintaining dim cockpit conditions in flight scenarios with flash blindness risk.
Metcalf and Horn conducted a flash blindness study that measured recovery times for the detection of slowly flashing adaptometer stimuli with luminances of 0.07, 0.45, 7, or 71 foot-lamberts (0.24, 1.5, 24, or 243 cd · m−2) following 100-ms carbon arc searchlight flashes ranging from 60 to over 12,000 lumens/ft2 (5.66 to over 7.95 logTd·s), after preadaptation to 0.7 foot-lambert (2.40 cd · m−2) brightness prior to flash exposure.15 They reported that recovery times increased apparently linearly with the log of the flash luminance and increased with lower stimulus luminance. These findings are consistent with ours, especially for the reflective attitude gauge. Though our stimuli and task differed from those in that study, our recovery times are roughly in similar proportion. For example, they reported a recovery time of approximately 35 s for a 0.45 foot-lambert (1.5 cd · m−2) stimulus after a flash of 3.79 log lumens/ft2 (7.67 logTd·s), while recovery time for our 1 cd · m−2 ambient luminance reflective gauge after a 7.5 logTd·s flash was approximately 30 s.
We chose the attitude task because of its importance for flying an aircraft successfully and because we thought the ability to perform this task, unlike locating the gap in a Landolt C target, relied on different visual characteristics. The horizon on an attitude indicator is a relatively large visual feature and the task of keeping it horizontal is an orientation discrimination task. That performing this task is likely easier than detecting the gap in a Landolt C is supported by generally faster recovery times after a flash. In the Landolt C task, recovery times varied from a few seconds to a couple of minutes depending on the size of the C, flash intensity, ambient luminance, and instrument type (i.e., reflective or emissive), compared to 6–30 s for the attitude task. It is less clear what visual cues subjects relied on to maintain balanced pitch and roll, which was the task after the initial level horizon recovery. Several cues are available, including the balance between ground and air and the pitch ladder and symbology associated with it. Since the subjects were not trained aviators, they may not have understood use of the pitch ladder and simply used whatever cues worked best during practice. Regardless, the pitch and roll error data suggest the difficulty of performing both functions was similar as the magnitude of the errors is similar.
The attitude indicator task also allowed us to assess longer term effects of flash exposure by tracking the ability to maintain straight-and-level flight over the course of 30 s when perturbations were introduced. For both the reflective and emissive attitude indicators, roll error increased with increases in flash intensity, but for pitch errors this was true only for the reflective attitude indicator. Thus, the roll error results are consistent with our hypothesis that tracking errors would increase with increases in flash intensity, but only the reflective attitude indicator results support this hypothesis for pitch errors. We also hypothesized that lower ambient luminance would increase tracking error after a flash for the reflective, but not the emissive, instrument type and both the pitch and roll error data supported this. We found that both pitch and roll errors were significantly larger at 1 cd · m−2 than 10 cd · m−2, but errors at 10 cd · m−2 were not larger than at 100 cd · m−2. The straight-and-level task could be regarded as involving a hyperacuity visual task because, similar to Vernier acuity stimuli, it involves detection of fine misalignments or offsets of lines, such as when the aircraft icon is just misaligned with the horizon line or when the icon is slightly translated relative to the horizon line.16 Hyperacuity involves visual detection with precision finer than the size of retinal photoreceptors and so implies that neural mechanisms beyond the optics of the eye underly this ability.17 Hyperacuity is more resilient to visual degradation, such as blur,18 so it may also be more resilient to flash blindness, though more research would be required to determine this.
A comparison of baseline tracking where no flash was present indicated that the ambient luminance factor did not act alone in modulating performance in our study. One-way ANOVAs conducted for each instrument type and tracking error type yielded no significant main effects of ambient luminance, suggesting the combined effects of flash and ambient luminance caused the increased errors at 1 cd · m−2 compared with 10 and 100 cd · m−2 for all reflective attitude indicator flash conditions. For the reflective attitude indicator, both pitch and roll errors were elevated compared to baseline for the 1 cd · m−2 ambient luminance condition for all flash conditions but not for any other combination of ambient luminance and flash intensity. In contrast, for the emissive attitude indicator, roll errors were elevated for the highest flash intensity for the 10 and 100 cd · m−2 ambient luminance conditions and nearly significant for the 1 cd · m−2 ambient luminance condition, but not for any other flash and ambient combination. Pitch errors with the emissive attitude indicator followed a different pattern with the significant increases in errors occurring with the two highest flash intensities in the lowest ambient luminance condition and for the intermediate flash intensity in the highest ambient luminance condition. It is not clear why roll and pitch errors followed different patterns for the emissive attitude indicator but showed the same pattern with the reflective attitude indicator. Despite the difference, most of the paired comparisons for the reflective and emissive attitude indicators are consistent with the inference that ambient luminance combined with a flash modulated tracking performance rather than ambient luminance alone.
In our previous study we concluded that our design isolated the effects of ambient luminance in the absence of glare. In the present experiment the viewing conditions were set the same way and we reached the same conclusion. The reason being the viewing conditions were set so the observer was not near the angle of reflection from the light source providing the ambient luminance. As a result, they did not experience an effect like reading a cell phone display in sunlight, where the sunlight can be reflected off the display surface and into the user’s eyes, making any display with a reflective glass overlay difficult to use. That does not mean viewing displays with glass overlays in an actual cockpit might not be affected by significant glare. In an actual cockpit under flight conditions, significant glare from reflection off glass overlays is potentially a problem as the reflected light could reduce visual contrast of the symbology and substantially increase flash blindness recovery time.
In summary, our results are consistent with a previous experiment using the same flash and ambient luminance parameters but a different spatial resolution task.11 Now for a dynamic tracking task we find, as in that study, that a reflective instrument is more advantageous for flash blindness recovery in high (100 cd · m−2) ambient luminance and the emissive instrument is more advantageous in low (1 cd · m−2) ambient luminance. The results of this experiment and our previous one will be used to inform pilot training procedures, develop a model of flash blindness recovery, and help define requirements for protection technologies such as laser eye protection and active tinting visors.
Reflective attitude indicator (left) and emissive attitude indicator created in unity (right) illuminated by light towers (not pictured). Photograph taken from subject’s vantage point.
Recovery times for the reflective attitude indicator (left) and the emissive attitude indicator (right). Error bars represent standard error of the mean.
Average roll error (top panels) and pitch error (bottom panels) vs. flash intensity for the reflective attitude indicator (left column) and the emissive attitude indicator (right column). Error bars represent standard error of the mean.
Average roll error (top row) and pitch error (bottom row) vs. ambient luminance level for the reflective attitude indicator (left column) and emissive attitude indicator (right column). The no-flash baselines are included for comparison against the flash conditions. Error bars represent standard error of the mean.
Contributor Notes
