Objectives To determine how central field loss (CFL) affects reaction time to pedestrians and to test the hypothesis that scotomas lateral to the preferred retinal locus will delay detection of hazards approaching from that side.
Methods Participants with binocular CFL (scotoma diameter, 7°-25°; visual acuity, 0.3-1.0 logMAR) using lateral preferred retinal fixation loci and matched controls with normal vision drove in a simulator for approximately 1½ hours per session for 2 sessions a week apart. Participants responded to frequent virtual pedestrians who appeared on either the left or right sides and approached the participant's lane on a collision trajectory that, therefore, caused them to remain in approximately the same area of the visual field.
Results The study included 11 individuals with CFL and 11 controls with normal vision. The CFL participants had more detection failures for pedestrians who appeared in areas of visual field loss than did controls in corresponding areas (6.4% vs 0.2%). Furthermore, the CFL participants reacted more slowly to pedestrians in blind than nonscotomatous areas (4.28 vs 2.43 seconds, P < .001) and overall had more late and missed responses than controls (29% vs 3%, P < .001). Scotoma size and contrast sensitivity predicted outcomes in blind and seeing areas, respectively. Visual acuity was not correlated with response measures.
Conclusions In addition to causing visual acuity and contrast sensitivity loss, the central scotoma per se delayed hazard detection even though small eye movements could potentially compensate for the loss. Responses in nonscotomatous areas were also delayed, although to a lesser extent, possibly because of the eccentricity of fixation. Our findings will help practitioners advise patients with CFL about specific difficulties they may face when driving.
Central field loss (CFL) is a scotoma that encompasses the fovea and is commonly caused by age-related macular degeneration; however, many other causes are possible.1,2 People with CFL almost always use a preferred retinal locus (PRL),3,4 an extrafoveal location near the scotoma, to fixate targets that would normally be foveally fixated (we will refer to scotoma location/direction relative to the PRL in visual field space, not in retinal directions). The scotoma is lateral to the PRL in approximately 65% of cases but can be above or, rarely, below the PRL.4,5 In addition, CFL reduces visual acuity (VA) and contrast sensitivity because these functions are normally poorer in the peripheral retina. In some countries (eg, the United Kingdom6 and Canada7), driving regulations address central visual field integrity and peripheral field extent. In the United States, however, driving regulations do not explicitly address CFL but rather consider only acuity loss.8 We hypothesize that vision loss due to CFL may have a greater effect on driving than simply acuity loss.
People with age-related macular degeneration report difficulty driving.9-13 However, many continue driving even when their VA falls below the legal limit and even when they have CFL.13,14 Delayed responses to stop signs and traffic lights have been reported for people with CFL in driving simulator studies.12,15 In an on-road study,16 25% of current drivers with age-related macular degeneration passed a driving test compared with 42% of people with peripheral field loss and 64% with other mild visual field impairments.
Although CFL is associated with driving difficulty, it is not known how the scotoma and its location affect driving skills. In a recent driving simulator study,17 people with hemianopia frequently failed to detect pedestrians appearing in their blind side of the road. We therefore hypothesized that scotomas lateral to the PRL would cause difficulty in detecting pedestrians appearing on that side despite the smaller size of the scotomas.
Visual acuity is widely used in driving regulations, but it is a poor predictor of performance.18 Contrast sensitivity is more predictive of driving outcomes in older adults with normal vision (NV)18 and is correlated with driving skills in people with moderate peripheral field loss.19 We therefore examined the relationship between pedestrian detection performance and a range of clinical vision measures, including scotoma size and location. We hypothesized that better contrast sensitivity and smaller scotoma size, but not better VA, would permit faster detection.
The study followed the tenets of the Declaration of Helsinki and was approved by institutional review boards at the Schepens Eye Research Institute and the Veterans Administration Boston Healthcare System.
Participants had at least a 120° horizontal binocular field extent, measured with Goldmann kinetic perimetry (V4e target). Corrected binocular single-letter VA was 20/200 or better for the CFL participants and 20/25 or better for NV controls. Thus, all had vision sufficient for a restricted drivers' license or better in some states.20 Each CFL participant had a binocular absolute central scotoma as measured with custom kinetic perimetry21 (74 candela-per-square-meter [cd/m2] bright 0.74° square target, gray background [24 cd/m2], distance of 1 m). Binocular scotoma location was categorized left or right of the binocular PRL in visual field space (ie, equivalent to a right PRL or left PRL, respectively). Individuals with PRLs above or below the scotoma were not included. A similar classification, based on the relative location of the PRL and former fovea, shows moderate repeatability (κ = 0.92 for 20 eyes of 12 participants) (Russell Woods, PhD, written communication, May 24, 2012).
Scotoma size was quantified as the mean diameter of 4 main meridians passing through the center of the scotoma. For one participant who had several distinct scotomas, each scotoma was measured and summed. Letter contrast sensitivity (2.5° letters) was measured with a custom, computer-based test with single-letter scoring, sequential decreasing contrast, and a 2-incorrect response stopping rule.22 The results are similar to those obtained with Pelli-Robson and Mars tests for patients with low vision.
Participants were recruited from the Veterans Administration, the Schepens Eye Research Institute, and the Harvard Cooperative Program on Aging. Participants with cognitive decline were excluded (>4 errors on the Short Portable Mental Status Questionnaire).23 All had more than 15 years of driving experience. None had previously used our simulator.
The simulator has been detailed previously.17,24 It is a PP1000-x5 simulator (FAAC Corp) with five 60 × 45-cm cathode ray tube monitors (1024 × 768 pixels, 60 Hz), providing a 225° × 32° field of view.
Two driving assessments were conducted roughly 1 week apart. Because of fatigue or discomfort, 5 participants completed assessments across more than 2 visits. Participants completed a series of acclimation and practice drives during which they rated their physical comfort and vehicle control on 10-point scales (“lousy” to “great”). If vehicle control was below 7, they continued to practice before progressing. Mean (SD) acclimation time was 18 (7) minutes.
Each assessment consisted of 3 city and 2 rural undivided highway scenarios, each lasting 8 to 12 minutes. We encouraged participants to drive 30 mph in the city and 60 mph on highways and to obey all standard road rules. Participants drove different scenarios during their first and second assessments, and 6 different counterbalance orders were used. Mean (SD) driving time for each session was 84 (11) minutes.
Participants pressed the horn as soon as they detected pedestrians, who appeared every 15 to 60 seconds (8-12 per scenario, 52 per session) at 1 of 4 eccentricities (−14°, −4°, 4°, and 14°). Pedestrians walked or ran (exhibiting biological motion) toward the road at a speed that would result in a collision with the car25,26 (eFigure 1). Thus, pedestrians stayed in approximately the same visual field location (eFigure 2), assuming the driver looked straight down the road. Although drivers may scan from side to side, even experienced drivers mainly look down the road in the direction of travel.27 Pedestrians stopped before entering the participant's lane.
Pedestrians appeared 67 m/134 m (city/highway) from the participant's vehicle. These distances are double the 2.5-second perception-brake time used in the calculation of minimum recommended sight distances for safe roadway design.28 At initial appearance, the pedestrians (2 m tall, light shirt, and dark pants) subtended 1.5° vertical and 0.5° horizontal in the city (half that on highways). Small eccentricities (−4° and 4°) represented pedestrians approaching from an adjacent lane (crossing the street) or the sidewalk. The larger eccentricities represented hazards approaching more quickly from a greater distance (eg, a bicyclist).
Primary measures were pedestrian detection rates and reaction times (latency from pedestrian appearance to horn-press). We used logistic regression to predict whether participants detected each pedestrian. Factors included visual field area in which each pedestrian appeared (ie, blind or seeing), drive type (city or highway), and vision (CFL or control). For CFL participants, whether pedestrians appeared in blind or seeing areas was based on the position and size of the scotoma in the binocular visual field plot (Figure 1). Visual field area (blind or seeing) was defined for controls by their matched CFL participant.
We analyzed median reaction times separately for (1) blind and seeing visual field areas, (2) drive type (city vs highway), and (3) first or second assessment. Medians were used because reaction times were not normally distributed. Medians did not include detection failures; these were used in the untimely reaction analysis. The medians were normally distributed and analyzed by repeated-measures analysis of variance, with area (blind or seeing), drive type, and assessment as within-subjects factors and vision (CFL or control) as a between-subjects factor (α=.05).
We calculated whether participants could have stopped in time, given their reaction time and vehicle speed, for each pedestrian appearance. A deceleration rate of 5 m/s2 was used, representing a car and road both in good condition.29 We classified each appearance as (1) timely, meaning the pedestrian was detected with enough time to stop if necessary, or (2) untimely, meaning the reaction was not quick enough to stop or the pedestrian was missed. Binary logistic regression was conducted with SPSS statistical software, version 11.5 (SPSS Inc), using backward stepwise entry based on significance of the Wald statistic.
We screened 28 individuals with CFL; 11 completed the study. Eight did not meet vision criteria and 9 withdrew: 2 for health reasons, 2 for simulator sickness, and 5 for other reasons (eg, transportation difficulties). For each CFL participant, a current driver with NV of the same sex and age (within 3 years) was recruited. We screened 17; 11 completed testing and were matched to a CFL participant.
Eight CFL participants had binocular scotomas to the right of their PRL, 2 to the left, and 1 to both the left and right (Figure 1), with scotoma diameters ranging from 7° to 25°. The CFL participants had poorer VA and contrast sensitivity than the NV participants (Table 1), and the 2 groups were similar for sex and age (Table 1).
Overall detection rates were high (Figure 1). The CFL participants had more detection failures than controls (2.7% vs 0.3%, Wald statistic = 14.8, df = 1, P < .001, Exp(β) = 10.3) and 2.1 times more misses for pedestrians in blind than seeing areas, which in turn was many times more than controls' corresponding areas (6.4% vs 0.2%, Wald statistic = 19.4, df = 1, P < .001, Exp(β) = 5.24). Drive type (city vs highway) was not significant (P = .20), and neither were any interactions.
Participants reacted 0.16 second faster at the second assessment, but this difference was not significant (P = .08). Overall, the CFL participants reacted significantly slower than controls (3.35 vs 1.27 seconds, F1,20 = 72.5, P < .001) (Figure 2) in both seeing and blind areas. As expected, the CFL participants reacted faster in seeing than in blind areas (2.43 vs 4.28 seconds, F1,21 = 50.4, P < .001), whereas in controls, reaction times did not differ (1.29 vs 1.25 seconds). For the CFL participants, the difference between seeing and blind areas was greater for rural highway than city drives (interaction of drive type [city vs highway] by area within the CFL participants, F1,21 = 9.3, P = .006). Controls did not differ by drive type.
The CFL participants were more likely to have untimely reactions than controls (29% vs 3%, Wald statistic = 44.44, df = 1, P < .001, Exp(β) = 0.02) (Figure 3). These untimely reactions were more likely to involve pedestrians in blind than in seeing areas (50% vs 19%) for CFL participants but not controls (5% vs 5%) (vision by area interaction, Wald statistic = 7.37, df = 1, P = .007, Exp(β) = 7.1) (Figure 3). All participants had more untimely reactions in highway than in city drives (48% vs 21% for CFL participants and 8% vs 1% for controls, Wald statistic = 9.89, df = 1, P = .002, Exp(β) = 0.12).
Vision measures and detection performance
Larger scotomas were correlated with lower detection rates and more untimely reactions for pedestrians in blind areas on city drives (Table 2). Poorer contrast sensitivity significantly correlated with longer reaction times and more untimely reactions in seeing areas on highways and worse detection rates in blind areas on city and highway drives. Age and VA were not correlated with the response measures. The multiple planned comparisons were not corrected.30,31
Our hypothesis that lateral CFL delays reactions to pedestrian targets in scotoma areas was strongly supported. Participants with scotomas (regardless of right or left PRL) had longer reaction times to pedestrians appearing in their blind areas than in their seeing areas. One participant with scotomas on both sides had delays on both except for the −4° targets, where there was residual vision. Despite the relatively small sample, our repeated measures of hazards at multiple eccentricities were sufficiently powerful to produce significant large median reaction time differences. Although our sample is unbalanced (8 with right CFL and 2 with left CFL), the proportions are close to those reported in a larger sample.4
The longer reaction times in blind areas were due to the scotoma and not simply to the loss of acuity and contrast sensitivity. Such large scotoma effects might not be expected because small eye movements might be sufficient to compensate for obscuration by a scotoma. However, in our sample, such scanning, if it took place, was not sufficient for full compensation.
The effects of CFL have been anticipated6 but have not been previously documented because of the difficulties of studying visually impaired driving. One on-road study of mild CFL32 used a “stunt” pedestrian and cyclist and found no apparent differences between people with CFL and controls in reaction times. This finding may be because (1) the timing of the actors could not be precisely implemented and (2) the authors stated that actors only appeared in seeing areas of the participants' visual fields. Thus, in that study, there was no ex post facto reason to have expected differences except those due to acuity or contrast sensitivity.
The CFL participants also had longer reactions than controls in seeing areas of their visual field. This finding might occur because most seeing-area pedestrians appeared at larger absolute retinal eccentricities for participants with CFL because they used nonfoveal PRLs, whereas controls fixated foveally. For example, a 4° pedestrian to a participant with a 6° PRL will be projected to 10° from the former fovea where contrast sensitivity is lower. This reduction in sensitivity caused more substantial effects at highway speeds (because detection needed to be made at a greater distance).
By comparison, in a study33 of drivers with paracentral scotomas who used foveal fixation and thus had VA and contrast sensitivity similar to that of NV drivers, no significant delays were found for pedestrian figures in seeing areas of their visual field using the same methods. Supporting the retinal eccentricity hypothesis, NV participants had longer reaction times to pedestrians at the larger (14°) eccentricities than the small (4°) by 0.3 second (paired t10 = 5.9, P < .001).
Despite deploying pedestrians at double the perception-brake sight distance in the American Association of State Highway and Transportation Officials guidelines28 (2 × 2.5-second travel time), the CFL participants frequently did not react in a timely fashion, especially in rural highway drives (69% untimely in blind areas and 28% in seeing areas). Timely detections on highways were challenging even for normally sighted participants (8% untimely) because doubling speed quadruples stopping distance.
Our primary measure was how quickly the pedestrian was detected. We also derived the measure “untimely reactions,” including vehicle speed and distance to the pedestrian, which imparts more real-world meaning to the measurement. We did not measure actual collisions, and we did not consider other possible maneuvers to avoid collisions. An advantage of simulator-based studies is that pedestrian challenges are safe, controlled, and more frequent than in on-road studies, enabling reliable measurement of response latencies. The greater frequency in the simulator should have primed the participants, making it easier for them to anticipate the events. In the real world, such occurrences would be unexpected and therefore would probably result in longer reaction times. It is not uncommon in mobility research to include events at a higher frequency or higher density than in the real world to have sufficient events for analysis.34,35
Larger scotomas were significantly correlated with poorer blind-area detection performance; larger blind areas should make it more difficult to detect pedestrians in that area. That contrast sensitivity correlated with worse detection rates in blind areas may be attributable to its correlation with scotoma size. For the acuity range of our CFL participants (20/40 to 20/200), VA was uncorrelated with performance measures despite being the primary vision screening measure for licensing in the United States. Higgins36 has pointed out that VA should not be expected to correlate with outcome measures when there are range restrictions. We should note that within our CFL or NV group, VA was not correlated with performance measures. However, across all our participants, VA was correlated with most performance measures.
Our CFL participants had vision sufficient for a restricted driver's license in some states but not in the United Kingdom or Canada because of their CFL.6,7 Although most had stopped driving, all had considerable driving experience, and the 3 who were current drivers had blind-area reaction times similar to the others.
In conclusion, people who fixate lateral to a binocular scotoma had relatively late reactions to potential hazards that appeared in scotoma locations. Vertical PRLs, more common in juvenile macular degeneration, may have a lesser effect on hazard detection, but confirmation is needed. Contrast sensitivity may also help differentiate those who are fit to drive. However, none of these measures is currently considered in driver licensing in the United States. We found that CFL may affect driving safety independent of its effect on acuity; thus, patients with CFL may be more vulnerable to hazards than other drivers with reduced acuity alone.
Our study was not designed to oppose or advocate for people with visual impairments as drivers. However, knowledge about how specific aspects of vision loss (CFL, VA, and contrast sensitivity) affect certain aspects of performance should help improve vision rehabilitation and the design of mobility aids. The results may help practitioners in advising patients with CFL about difficulties they may face when driving.
Correspondence: P. Matthew Bronstad, PhD, Schepens Eye Research Institute, 20 Staniford St, Boston, MA 02114 (matthew.bronstad@schepens.harvard.edu).
Submitted for Publication: March 23, 2012; accepted September 11, 2012.
Published Online: January 17, 2013. doi:10.1001/jamaophthalmol.2013.1443
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by grants EY12890 (Dr Peli), EY018680 (Dr Bowers), and 2P30EY003790 from the National Institutes of Health.
Additional Contributions: Christina Gambacorta, BA, Egor Ananev, BS, and Alex Hwang, PhD, made helpful comments on the manuscript. Joseph Rizzo, MD (Center for Innovative Visual Rehabilitation, Veterans Administration Boston Healthcare System), provided the simulator. The Harvard Cooperative Program on Aging assisted in recruitment.
1.Sunness JS, Applegate CA, Haselwood D, Rubin GS. Fixation patterns and reading rates in eyes with central scotomas from advanced atrophic age-related macular degeneration and Stargardt disease.
Ophthalmology. 1996;103(9):1458-14668841306
PubMedGoogle Scholar 2.Petzold A, Plant GT. Central and paracentral visual field defects and driving abilities.
Ophthalmologica. 2005;219(4):191-20116088237
PubMedGoogle ScholarCrossref 3.Timberlake GT, Peli E, Essock EA, Augliere RA. Reading with a macular scotoma, II: retinal locus for scanning text.
Invest Ophthalmol Vis Sci. 1987;28(8):1268-12743610545
PubMedGoogle Scholar 4.Verezen CA, Hoyng CB, Meulendijks CF, Keunen JE, Klevering BJ. Eccentric gaze direction in patients with central field loss.
Optom Vis Sci. 2011;88(10):1164-117121804438
PubMedGoogle ScholarCrossref 5.Fletcher DC, Schuchard RA. Preferred retinal loci relationship to macular scotomas in a low-vision population.
Ophthalmology. 1997;104(4):632-6389111255
PubMedGoogle Scholar 6.Rauscher FG, Chisholm CM, Crabb DP,
et al. Central Scotomata and Driving. London, England: Department of Transport; 2007
8.Peli E. Low vision driving in the USA: who, where, when, and why.
CE Optometry. 2002;5:54-58
Google Scholar 9.Mangione CM, Berry S, Spritzer K,
et al. Identifying the content area for the 51-item National Eye Institute Visual Function Questionnaire.
Arch Ophthalmol. 1998;116(2):227-2339488276
PubMedGoogle Scholar 10.Scilley K, Jackson GR, Cideciyan AV, Maguire MG, Jacobson SG, Owsley C. Early age-related maculopathy and self-reported visual difficulty in daily life.
Ophthalmology. 2002;109(7):1235-124212093644
PubMedGoogle ScholarCrossref 11.Ball K, Owsley C, Stalvey B, Roenker DL, Sloane ME, Graves M. Driving avoidance and functional impairment in older drivers.
Accid Anal Prev. 1998;30(3):313-3229663290
PubMedGoogle ScholarCrossref 12.Szlyk JP, Pizzimenti CE, Fishman GA,
et al. A comparison of driving in older subjects with and without age-related macular degeneration.
Arch Ophthalmol. 1995;113(8):1033-10407639654
PubMedGoogle ScholarCrossref 13.DeCarlo DK, Scilley K, Wells J, Owsley C. Driving habits and health-related quality of life in patients with age-related maculopathy.
Optom Vis Sci. 2003;80(3):207-21312637832
PubMedGoogle ScholarCrossref 14.Bowers AR, Apfelbaum DH, DeCarlo D, Peli E. Use of bioptic telescopes by drivers with age-related macular degeneration [abstract].
Invest Ophthalmol Vis Sci. 2007;48:2351
Google Scholar 15.Szlyk JP, Fishman GA, Severing K, Alexander KR, Viana M. Evaluation of driving performance in patients with juvenile macular dystrophies.
Arch Ophthalmol. 1993;111(2):207-2128431157
PubMedGoogle ScholarCrossref 16.Coeckelbergh TRM, Brouwer WH, Cornelissen FW, Kooijman AC. Predicting practical fitness to drive in drivers with visual field defects caused by ocular pathology.
Hum Factors. 2004;46(4):748-76015709335
PubMedGoogle ScholarCrossref 17.Bowers AR, Mandel AJ, Goldstein RB, Peli E. Driving with hemianopia, I: detection performance in a driving simulator.
Invest Ophthalmol Vis Sci. 2009;50(11):5137-514719608541
PubMedGoogle ScholarCrossref 19.Bowers AR, Peli E, Elgin J, McGwin G Jr, Owsley C. On-road driving with moderate visual field loss.
Optom Vis Sci. 2005;82(8):657-66716127330
PubMedGoogle ScholarCrossref 20.Peli E, Peli D. Driving With Confidence: A Practical Guide to Driving With Low Vision. Toh Tuck Link, Singapore: World Scientific Publishing Co Pte Ltd; 2002
22.Arditi A. Improving the design of the Letter Contrast Sensitivity Test.
Invest Ophthalmol Vis Sci. 2005;46(6):2225-222915914645
PubMedGoogle ScholarCrossref 23.Pfeiffer E. A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients.
J Am Geriatr Soc. 1975;23(10):433-4411159263
PubMedGoogle Scholar 24.Peli E, Bowers AR, Mandel AJ, Higgins K, Goldstein RB, Bobrow L. Design of driving simulator performance evaluations for driving with vision impairments and visual aids.
Transport Res Rec. 2005;1937:128-135
Google ScholarCrossref 26.Chardenon A, Montagne G, Buekers MJ, Laurent M. The visual control of ball interception during human locomotion.
Neurosci Lett. 2002;334(1):13-1612431764
PubMedGoogle ScholarCrossref 28.American Association of State Highway and Transportation Officials. A Policy on Geometric Design of Highways and Streets. Washington, DC: American Association of State Highway and Transportation Officials; 2004
29.Evans L. Traffic Safety. Bloomfield Hills, MI: Science Serving Society; 2004
32.Lamble D, Summala H, Hyvärinen L. Driving performance of drivers with impaired central visual field acuity.
Accid Anal Prev. 2002;34(5):711-71612214966
PubMedGoogle ScholarCrossref 33.Bronstad PM, Bowers AR, Albu A, Goldstein RG, Peli E. Hazard detection by drivers with paracentral homonymous field loss: a small case series.
J Clin Exp Ophthalmol. 2011;S5:001
Google Scholar 34.Fuhr PS, Liu L, Kuyk TK. Relationships between feature search and mobility performance in persons with severe visual impairment.
Optom Vis Sci. 2007;84(5):393-40017502822
PubMedGoogle ScholarCrossref 35.Wood JM, Chaparro A, Lacherez P, Hickson L. Useful field of view predicts driving in the presence of distracters.
Optom Vis Sci. 2012;89(4):373-38122366710
PubMedGoogle ScholarCrossref 36.Higgins KE. Low vision driving among normally-sighted drivers. In: Rosenthal B, Cole R, eds. Remediation and Management of Low Vision. St Louis, MO: Mosby; 1996