The Effect of Simulated Central Field Loss on Street-Crossing Decision-Making in Young Adult Pedestrians

Abstract

Significance.

This study explored the street-crossing decision-making performance of young normally-sighted subjects with simulated central field loss. The results suggest that using eccentric viewing enables a person to make safe and reliable street-crossing decisions.

Purpose.

This study tested the hypothesis that as the diameter of an experimentally induced central scotoma increases, the accuracy and reliability of street-crossing decisions worsen.

Methods.

Street-crossing decisions were measured in twenty young subjects aged between 23 and 31 years while monocularly viewing a non-signalized, one-way street for different vehicular arrival times. Using a 5-point rating scale, subjects judged whether they could cross the street prior to vehicular arrival with habitual vision and simulated central field loss (CFL) with eccentric viewing. The CFL was induced using soft contact lenses with different central opaque diameters. Using Receiver Operating Characteristics (ROC) curve analysis, we obtained subjects’ accuracy (amount of time in seconds where subjects either over-or under-estimated vehicular arrival time relative to their actual crossing time) and reliability (how quickly subjects transitioned from judging insufficient to sufficient time to cross relative to their actual crossing time).

Results.

The centrally opaque contact lenses induced central scotomata with mean (standard deviation) diameter of 17.12° (5.83°). No significant difference in street-crossing accuracy (P = .35) or reliability (P =.09) were found between the normal, habitual vision and simulated CFL conditions. No statistically significant correlations were found between scotoma diameter and the accuracy and reliability of subjects’ street-crossing decisions (P = .83 and P = 0.95, respectively).

Conclusions.

The findings of this study suggest that adopting eccentric viewing enables a person to successfully mitigate the negative effects of an absolute central scotoma on the accuracy and reliability of their street-crossing decisions.

The fovea is the specialized central area of the macula, and it provides the highest level of visual acuity since it contains the maximum density of cone photoreceptors.¹ As a result, normally-sighted individuals fixate with their fovea when performing various activities of daily living such as watching television² and reading.³ There are ocular diseases however, that result in a progressive degeneration of the macula including the fovea which results in central vision loss. Age-related macular degeneration is the leading cause of central vision loss in people aged 50 and older in western countries.⁴ Stargardt’s Disease and Degenerative Myopia can also cause central vision loss but early in life-typically during the teenage years or as an early adult.^5,6 The central vision loss experienced by people with these macular degenerative diseases is characterized by reduction in visual acuity, contrast sensitivity, and by central field loss.^5–8

With compromised foveal vision, individuals with central vision loss often use a relatively healthy, non-foveal retinal location to fixate and perform visual tasks. This adaptive strategy is called eccentric viewing and the selected non-foveal retinal location used for fixation is termed the preferred retinal locus.^{9, 10} However, numerous studies have shown that the majority of patients with central field loss are unaware of the presence of their central scotomata¹¹ which makes it difficult for them to maintain eccentric viewing and thus effectively use their preferred retinal locus.^12,13

Despite long-term use of eccentric viewing, individuals with central vision loss still have difficulties when performing various activities of daily living such as recognizing faces, reading and driving.^14–16 Given that driving is the main method used for commuting in western countries,¹⁷ those individuals who cease driving need to rely on alternative transportation modes (such as public transportation, taxis or lifts from relatives and friends) to maintain their independence. Driving cessation may also result in an increase in walking as a means of personal transportation. Patients with central vision loss report having difficulty with mobility,¹⁸ and it has been shown that mobility performance becomes significantly less efficient as the size of the binocular central field loss increases.¹⁹ This reduced mobility performance in the presence of central field loss might therefore also significantly affect a person’s ability to perform the high risk and complex mobility task of crossing the street.

Crossing the street is a common activity of daily life. Pedestrians are among the most vulnerable users of the road, and an inappropriate crossing decision may result in injury or even loss of life. According to the World Health Organization, more than 275,000 pedestrians are killed globally each year accounting for 22% of all road traffic deaths worldwide.²⁰ Pedestrian crashes affect people from all different age groups. In the United States in 2017, pedestrians aged 65 years and older accounted for 20% of all traffic fatalities while people aged between 20 and 34 years accounted for 12% of all traffic fatalities.²¹ In Australia in 2017, pedestrians aged between 17 and 39 years accounted for 22% of pedestrian fatalities,²² whereas 64% of pedestrians involved in road traffic crashes in India in 2017 were aged between 18 and 45 years.²³ Whilst the aforementioned statistics of pedestrian fatalities are not specific to visually impaired pedestrians and do not differentiate between fatalities owing to pedestrian or driver error /carelessness, these high percentages of fatalities demonstrate that pedestrian safety is an important global concern that impacts society across the entire age spectrum.

The effect of central vision loss on street-crossing performance has been assessed in only a few studies.^24–28 In three of these earlier central vision loss street-crossing studies,^24–26 subjects only had reduced visual acuity and contrast sensitivity and no central field loss (i.e. absolute scotoma). In the remaining two studies^{27, 28} subjects had a central field loss in addition to reduced visual acuity and contrast sensitivity.

The majority of these earlier studies have shown that the street-crossing performance of subjects with central vision loss, including those with central field loss, was comparable to that of age-matched normally-sighted subjects.^24–27 However, Geruschat et al.²⁴ found that age-related macular degeneration subjects with central vision loss but without a central field loss, took significantly longer to identify crossable vehicular arrival times and had significantly larger negative (unsafe) safety margins compared to the age-matched normally-sighted subjects.

No studies however have assessed the street-crossing decision-making performance of young individuals with central field loss in real outdoor environments. Wu et al.²⁸ studied crossing decisions of young normally-sighted subjects simulated with a gaze-contingent central field loss within a virtual environment. They found that with increasing scotoma size, subjects had longer curb delay and selected longer time gaps between traffic when crossing an exit lane of a virtual roundabout. However, this behavior was expected since their subjects were required to fixate centrally. Thus as the scotoma size became larger, subjects had to wait increasingly longer until the angular subtense of the approaching vehicle was greater than that of the central scotoma. Additionally, while their virtual environment was immersive and contained 3-D auditory cues, it still may not have resulted in behaviorally equivalent performance as observed in the real-world.

Given that young people as a whole are involved in pedestrian fatalities and that they walk significantly more than older people,²⁹ they may encounter the task of street-crossing, and hence its associated risks and dangers, more frequently than older people when they travel to school, work and/or exercise. The aim of the present study was to determine the effects of simulated central field loss and eccentric viewing on street-crossing decision-making performance in a group of young adult pedestrians along a real-world street. The central field loss was induced using a gaze-contingent experimental model of central field loss.³⁰

We hypothesized that normally-sighted young subjects using eccentric viewing under the simulated central field loss condition would adopt a risk averse street-crossing strategy compared to the habitual vision condition. This is because with the elimination of their central vision for the first time, subjects may become more cautious by classifying only those vehicular gap times that are significantly longer in duration than their actual crossing time as being “enough time to cross”. Adopting such a strategy will result in subjects to become more inaccurate with their decision-making under the central field loss condition compared to when using their normal central (foveal) vision.

Additionally, when maintaining eccentric viewing, subjects will use a preferred retinal locus that is most likely located within the peripheral visual field. Given that vehicular gap time judgments are more variable when using the peripheral visual field than when using foveal (central) vision,³¹ we hypothesize that subjects will be more variable and hence less reliable in their street-crossing decisions when using eccentric viewing (peripheral visual field) compared to when using their normal central vision.

METHODS

Subjects

Twenty-four normally-sighted subjects aged between 23 to 31 years (average (standard deviation (SD)) = 26.98 (2.23) years) participated in the study. Inclusion criteria included age between 18 and 40 years, full cognitive and physical function, the ability to walk and stand unaided, corrected right eye visual acuity of 20/30 or better, absence of any ocular diseases resulting in vision loss, and good general health. Subjects were recruited from friends and students at the Indiana University School of Optometry after obtaining informed consent. The present study adhered to the tenets of the Declaration of Helsinki, and was approved by the Institutional Review Board of Indiana University.

Visual Function Measures

Habitual Vision Condition

Monocular, right eye visual acuity, contrast sensitivity, and visual field were measured in each subject while the left eye was occluded. The visual acuity and contrast sensitivity were assessed while subjects wore their habitual spectacle prescription. Visual acuity was measured using a Lighthouse Early Treatment Diabetic Retinopathy Study acuity chart that was trans-illuminated to approximately 85 candela/meter² and reported as the Logarithm of the Minimum Angle of Resolution (logMAR) using the scoring of Bailey and Lovie.³² Contrast sensitivity was measured at 1 meter using an Evans Letter Contrast Sensitivity Chart trans-illuminated to approximately 85 candela/meter². Subjects were assigned Log CS scores using the scoring of Elliott et al.^33,34

Monocular, right eye semi-automated kinetic perimetry was also performed on each subject using the Octopus 900 perimeter (Haag Streit, Ohio, USA). To avoid lens rim effects, subjects did not wear their habitual spectacle prescription during the visual field assessment. Subjects were instructed to fixate a center point while a III4e Goldmann stimulus (0.43° stimulus diameter, 0 decibel intensity) was presented against a 10 candela/meter² background luminance level along 12 meridians separated by 30° intervals. The light stimulus was moved from the far periphery towards the perimeter bowl center. Subjects were instructed to press a button whenever they first saw the light stimulus moving in their peripheral visual field. This corresponded to their visual field extent. To obtain the borders of the blind spot or any scotomata, subjects were instructed to press the button if the light disappeared as it moved from the visual field extent towards the center of the bowl and press the button again if the light reappeared. These two locations corresponded to the outer and inner edge of the blind spot or scotoma, respectively. The blind spot or any scotoma were then mapped by moving the stimulus at a rate of 5º/second from the blind spot or scotoma center (non-seeing) along the eight principal meridians until the subject reported when they first saw the stimulus (seeing). For each subject, visual field extent (radius) was averaged across the 12 meridians, and the visual field area was computed as the area contained within the visual field extent. The remaining visual field area was calculated by subtracting the combined area of the blind spot and any scotoma from the area contained within the visual field extent. As expected, no central scotomata were detected under the habitual vision condition.

Simulated Central Field Loss Condition

Each subjects’ right eye was fitted with a centrally-opaque soft contact lens using the methods described in Almutleb et al.³⁰ In summary, each subject’s right eye was fitted with a prism ballast, plano, centrally-opaque soft contact lens with a central opacity diameter of either 2.8, 3.0 or 3.2 mm. Only one central-opacity contact lens was fitted for each subject for a total of 24 eye-contact lens pairings.

With the left eye occluded, right eye visual acuity at the street test site was measured using the Early Treatment Diabetic Retinopathy Study acuity chart while subjects maintained eccentric viewing that resulted in the best vision for them (central field loss visual acuity with eccentric viewing). The central field loss visual acuity with eccentric viewing was reported as the Log MAR and scored using the methods of Bailey and Lovie.³²

With the left eye occluded, the extent and position of the induced central, absolute scotoma in the right eye was assessed at the street test site using a custom-made Tangent (Bjerrum) Screen positioned at 0.90 meter and a custom-made 5 mm white Traquair target. The Tangent (Bjerrum) Screen had a reflectance of 28% (gray background) which was similar to the average reflectance of the environment at the test street (i.e. average combined reflectance of grass, trees, aged-asphalt, and black, white and colored vehicles).^{35, 36} To ensure that the contact lens opacity obscured the fovea, thus resulting in an absolute central field loss, subjects were instructed to centrally fixate on a large central fixation target on the Tangent (Bjerrum) Screen. Being able to fixate the central target meant that the contact lens opacity was decentered to such an extent that the resulting displacement of the scotoma enabled foveal fixation. In these cases, subjects were excluded from the study. In this study, four out of the 24 subjects were excluded because of their ability to centrally fixate while wearing the centrally-opaque contact lens. As a result, data were collected on a total of 20 eye-contact lens pairings.

When subjects were unable to see the fixation target on the Tangent (Bjerrum) Screen, they were instructed to maintain eccentric viewing by placing one edge of the scotoma inferior, superior, nasal or temporal to the fixation target on the Tangent (Bjerrum) Screen. Subjects selected which eccentric viewing direction resulted in the best vision for them. Scotoma extent was then mapped from non-seeing to seeing following standard clinical practice.^{37, 38} Scotoma Diameter in degrees was obtained for each subject by averaging the scotoma radius across the eight principal meridians and multiplying this result by two. The centrally opaque contact lenses induced central, absolute scotomata that ranged from 6.32° to 28.20° in diameter with a mean (SD) diameter of 17.12° (5.83°).

Cognitive and Health Assessments

Cognitive functions were also assessed in each subject using the Mini-Mental State Exam, and the Trails Making Test – Part B. Refer to Table 1 for the average (SD) Mini-Mental State Exam and Trails Making Test – Part B scores for all subjects.

Table 1.

Mean (standard deviation) of subjects’ cognitive function and general health assessment

Cognitive Assessment		General Health Assessment (Percentage)
		Physical Health				Mental Health
MMSE	TMT – Part B (seconds)	Physical Functioning	Role-Physical	Bodily Pain	General Health	Vitality	Social Functioning	Role-Emotional	Mental Health
29.85 (0.67)	38.88 (13.56)	99.50 (2.24)	98.75 (5.60)	89.80 (14.78)	83.85 (10.96)	71 (14.10)	94.38 (9.50)	91.67 (26.21)	84.40 (5.93)

MMSE = Mini-Mental State Exam, TMT – Part B = Trails Making Test – Part B. General health assessment included the eight domains from the Short Form 36 (SF-36) Health Questionnaire.

The health status of participating subjects was also assessed using the Short Form 36 Health Questionnaire, which measures physical health as well as mental health.³⁹ The Short Form 36 Health Questionnaire consists of eight domains (Table 1) and each domain has a score ranging between zero percent (poorest health) and 100% (best health).³⁹ Refer to Table 1 for subjects’ average (SD) score for each of the eight Short Form 36 domains.

Street-Crossing Survey

Subjects completed a survey about their street-crossing habits for non-signalized streets. Subjects answered whether they independently crossed a non-signalized street by selecting one of the following options: never required assistance, sometimes required assistance, always required assistance or never crossed a non-signalized street. Subjects also answered questions rating the frequency of crossing different types of streets (such as a one-way street, two-way street and roundabout) by selecting one of the following options: never, rarely (defined as a few times per month), often (defined as several times per week) or I do not encounter this type of street on foot. For questions rating the perceived difficulty of crossing each of the different types of streets, a five-point Likert scale from no difficulty (rating= 1) to impossible (rating= 5) was used. A five-point Likert scale from very conservative (rating= 1) to very liberal (rating= 5) was also used when subjects rated their own crossing behavior.

The Street Test Site

The test site used in the present study was located in Bloomington, IN and has been described previously.^{27, 40} In summary, the street was a non-signalized, two-way street with one lane of traffic approaching in either direction. A traffic island separated both lanes, and this allowed subjects to make street-crossing decisions about a single lane of traffic (approximately 4.62 meters wide) approaching from the lane closest to them. Subjects were seated by the curb, at a designated location called the “crossing point”, where they were required to make street-crossing decisions. This “crossing point” remained the same throughout the duration of the study.

Vehicular Arrival Time Measurements

Vehicular velocity and arrival times were measured using the methods described in previous studies.^25–27 In summary, two laser sensors were positioned on the curb of the test street. Laser sensor 1 was positioned in front of the subject at the crossing point and was defined as 0 meter while laser sensor 2 was approximately 2.0 meters to the left of the crossing point where the subject was sitting. Two retro-reflectors on the other side of the street were aligned with each of the laser sensors. Whenever a vehicle interrupted the laser beam, the event was recorded, timestamped and sent, via a USB cable, to a laptop computer which contained a custom-written “street-crossing” program. Vehicular speed was computed by dividing the known distance between the two sensors by the time it took the vehicle to travel between the two sensors. The physical vehicular arrival time was defined as the duration between the time of the prompt signal (prompt signal is explained in the Experimental Procedure section below) and the time when the first approaching vehicle interrupted the laser sensor positioned at the crossing point.

Experimental Procedure

The experimental procedure used to obtain street-crossing decisions and crossing times have been previously described in Hassan et al.^25–27 All street-crossing decisions and crossing times were measured monocularly with the right eye while the subject’s left eye was occluded. At the start of the study with their left eye occluded, subjects physically crossed the street at their usual walking pace four times under two viewing conditions (habitual vision and the simulated central field loss condition). This provided subjects with information about the time and distance required to physically cross the street. The order of measuring subjects’ actual crossing times under habitual vision and simulated central field loss conditions was counterbalanced across subjects. Subjects then sat by the side of the road at the crossing point and made street-crossing decisions under two conditions, habitual vision with central fixation and simulated central field loss with eccentric viewing, while monocularly viewing with their right eye the test street for different vehicular arrival times (0.1 — 19 seconds in duration).

Under the habitual vision condition, subjects were instructed to direct their head forward while wearing earbuds that played white noise. This was done to prevent subjects from sampling any visual and auditory information before they had to make their street-crossing decision. After an audible “get-ready” signal was given, the white noise stopped playing and subjects were instructed to view and listen to approaching traffic for two seconds. Because previous studies have shown that normally-sighted⁴¹ and subjects with central field loss²⁴ take on average up to two seconds to make a street-crossing decision, subjects were given two seconds to observe traffic from which to make their crossing decision. At the end of the two seconds, a “prompt” signal was automatically given and the white noise also automatically resumed playing. At the time of the prompt signal, subjects were instructed to immediately look straight ahead again, and were required to judge whether they believed that there was enough time to cross prior to vehicular arrival by clicking a button that was attached to a laptop computer.

Under the simulated central field loss condition, the same procedure was used as with the habitual vision condition except that subjects were instructed to maintain eccentric viewing by placing the edge of the induced scotoma to one side of the street just adjacent to the curb such that they could see approaching vehicles using their peripheral visual field. This is analogous to a patient using a preferred retinal locus at one of the edges of the scotoma. The scotoma/preferred retinal locus location used by subjects when making their street-crossing decisions was determined by the subject when their visual field was measured using the Tangent (Bjerrum) Screen. For example, if they placed the scotoma to the right of the fixation target (i.e. left-field preferred retinal locus), they would place the scotoma to the right side of the street just adjacent to the curb. Over half of the subjects (55%) used a left-field preferred retinal locus where 35% and 10% of the subjects used right-field and superior-field preferred retinal locus, respectively.

Trials where the approaching vehicle changed speed during the trial were discarded. The average (SD) speed for all trials and sessions under habitual vision and simulated central field loss conditions were 49.58 (1.64) kilometer/hour) and 49.24 (2.58) kilometer/hour), respectively. Therefore, approaching vehicles were on average slightly faster than the 48 kilometer/ hour speed limit of the street test site for all conditions and sessions. All subjects were instructed to make street-crossing decisions based on the assumption that the approaching vehicle would never slow down or yield to them.

Subjects used a 5-point rating scale as validated by Hassan and Massof²⁶ to indicate whether they believed that the vehicular arrival time was longer or shorter in duration than their crossing time (Table 2). Subjects pressed the button the same number of times corresponding to the desired rating number.

Table 2.

The 5-point rating scale that subjects used when making their street-crossing decisions.

Rating Category	Meaning
1	Definitely not enough time to cross: Vehicle arrival time is definitely shorter than subject’s crossing time
2	Probably not enough time to cross: Vehicle arrival time is somewhat shorter than subject’s crossing time
3	About the same time: Vehicle arrival time is approximately equal to the subject’s crossing time
4	Probably enough time to cross: Vehicle arrival time is somewhat longer than subject’s crossing time
5	Definitely enough time to cross: Vehicle arrival time is definitely longer than subject’s crossing time

Street-crossing decisions were collected for nine different vehicular arrival time categories (Table 3), and a minimum of ten trials were collected for each vehicular arrival time category. Before collecting street-crossing decision data for both viewing conditions, practice trials were given until the subject demonstrated understanding of the instructions regarding the experimental task, using the full rating scale and for the simulated central field loss condition how to maintain eccentric viewing. The testing order of the two viewing conditions (habitual and central field loss with eccentric viewing) was counter-balanced across subjects.

Table 3.

Vehicular arrival time categories for which street-crossing decisions were collected.

Vehicle Arrival Time (seconds)	Vehicle Arrival Time Category
0 – 0.99	1
1 – 1.99	2
2 – 2.99	3
3 – 3.99	4
4 – 4.99	5
5 – 5.99	6
6 – 6.99	7
7 – 7.99	8
≥ 8.00	9

Data Analysis

We analyzed the crossing decision data using the methods developed and validated by Hassan and Massof.²⁶ In summary, we computed the cumulative frequency distributions of subjects’ crossing decisions for each vehicular arrival time category by calculating the proportion of times that subjects used a specific rating for a given vehicular arrival time category. Graphing the cumulative frequency distribution of one vehicular arrival time category against the cumulative frequency distribution of another vehicular arrival time category, gave us the Receiver Operating Characteristic (ROC) curve for that vehicular arrival time pair. ROC curves were graphed for all vehicular arrival time pairs for each subject for both test conditions, resulting in a total of 36 ROC curves for each subject for each condition. Computing the area under each ROC curve and converting each area to a z score and multiplying this result by the square root of 2, we obtained a dis-similarity value (d’) for each vehicular arrival time pair. D prime (d’) represents the distance between the means of the frequency distributions of each vehicular arrival time pair. There were a total of 36 d’ values for each subject and condition. Entering the 36 d’ values into a one-dimensional scaling model, we estimated the means of the decision variable for each vehicular arrival time category for each subject and condition. Plotting these nine means as a function of vehicular arrival time categories for each subject and condition, we obtained the best fitting non-linear curve from which we obtained two parameters for each subject and condition: the x-intercept and the slope of the non-linear function at the x-intercept. The x-intercept represented the time in seconds when a subject classified vehicular arrival time from being “not enough time to cross” to being “enough time to cross”. Thus this point represented the transition point between perceived unsafe and safe vehicular arrival times.^25–27 The slope at the x-intercept represented the rate of change in subjects’ criterion for street-crossing decisions.^25–27

Calculating the difference between the x-intercept of the non-linear function and subjects’ averaged actual crossing time for the respective viewing condition, gave us each subject’s bias for each condition. Bias is the amount of time, in seconds, where subjects either over-or under-estimated vehicular arrival time relative to their actual crossing time.^25–27 A zero bias indicated perfect accuracy in street-crossing decisions. Positive bias values suggest inaccurate but safe street-crossing decision-making performance. This is because subjects would not classify vehicular arrival time as being “enough time to cross” until they were longer in duration than their actual crossing time. The converse is true for negative bias values.

The reliability of subjects’ street-crossing decision-making for each condition was determined as the slope of the non-linear function at the x-intercept.^25–27 Steep slopes indicated that the subjects were reliable in their street-crossing decisions because the subject transitioned quickly from judging vehicular arrival time as not enough to enough time to cross. The converse is true for subjects with shallow slopes.

Kolmogorov-Smirnov tests were performed to determine the normality of the distributions of bias, slope, vision measurements, cognition functions and health status assessments obtained under the habitual vision and/or simulated central field loss with eccentric viewing conditions. Parametric analyses were used with normally distributed data while non-parametric analyses were used on non-normally distributed data if it could not be transformed into a normal distribution.

To assess for accuracy in street-crossing decisions under both conditions (habitual vision and simulated central field loss with eccentric viewing), the two distributions of bias values were each tested if they were significantly different from zero using one sample sign tests. Wilcoxon signed ranks tests were performed to assess for significant differences in accuracy (bias) and reliability (slope) between habitual vision and simulated central field loss with eccentric viewing conditions. To assess if there was clustering in our data, we tested whether the distribution of difference scores (habitual – central field loss with eccentric viewing) for accuracy (bias) and reliability (slope) were normally distributed. Spearman’s rank-order correlations were performed to assess for any relationship between accuracy (bias) and reliability (slope) with measures of vision, cognition and health status under both viewing conditions (habitual and simulated central field loss with eccentric viewing).

Backward elimination multiple regression analyses were performed in two steps to determine the best predictor(s) of accuracy (bias) and reliability (slope) of street-crossing decision-making under habitual and simulated central field loss with eccentric viewing conditions. Four backward multiple regression analyses were initially performed for the two dependent variables (i.e. accuracy (bias) and reliability (slope)) under the two viewing conditions (i.e. habitual vision and simulated central field loss with eccentric viewing conditions). The independent variables (predictors) in these four multiple regression analyses were the eight domains of the Short Form-36 Health Questionnaire. This was done to reduce the number of independent variables (predictors) in the multiple regression analysis models given the small sample size. Another four multiple regression analyses were then performed for the same dependent variables under the two viewing conditions. The predictor variables under the habitual vision condition were habitual visual acuity, contrast sensitivity, visual field extent, visual field area, remaining visual field area, Mini-mental State Examination, Trails Making Test – Part B, and any significant Short Form-36 Health Questionnaire domains from the initial round of multiple regression analyses for each dependent variable (i.e. accuracy (bias) and reliability (slope)). For the simulated central field loss with eccentric viewing multiple regression analysis models, the same predictor variables as those used in the habitual vision condition multiple regression analyses were used except that habitual visual acuity and contrast sensitivity were excluded and central field loss visual acuity with eccentric viewing and scotoma diameter were included along with any significant Short Form-36 Health Questionnaire domains from the initial round of multiple regression analyses for each dependent variable. To enter a variable into the multiple regression analysis models, a P≤ .05 criterion was used and a P≥ .051 criterion was used to remove a predictor variable from the multiple regression analysis model. Assumptions about normality, homoscedasticity of residuals and multicollinearity were all examined in all multiple regression analyses.

RESULTS

Subject Characteristics

Under the habitual vision condition, the normally-sighted young subjects had average (SD) Log MAR visual acuity and Evans letter contrast sensitivity Log CS values of −0.14 (0.07) and 2.17 (0.10), respectively. Subjects’ unoccluded visual fields were within normal limits and had average (SD) visual field extent radius, area and remaining area of 63.12° (2.82°), 12,335.35 (1057.45) degrees² and 12,290.30 (1058.50) degrees², respectively.

Under the simulated central field loss with eccentric viewing condition, subjects’ average (SD) Log MAR visual acuity was 0.80 (0.25) which was significantly worse than the visual acuity under the habitual vision condition (Paired t-test, t(19) = −15.10, P < .0001).

The average (SD) results of the cognitive and general health assessments are shown in Table 1.

Accuracy and Reliability of Street-Crossing Decisions

Figure 1 shows the median bias values under habitual vision and simulated central field loss with eccentric viewing. No significant difference in bias was found between the two viewing conditions (Wilcoxon Rank Test Z = −0.93, P = .35, Figure 1). Bias values however under both conditions were on average significantly different from zero (one sample sign tests, P < .0001 for both habitual and central field loss with eccentric viewing conditions, Figure 1) but positive in value indicating that subjects on average made significantly inaccurate but safe street-crossing decisions under the habitual vision and the simulated central field loss with eccentric viewing conditions.

Figure 1.

Street-crossing accuracy (measured as Bias in seconds) under normal habitual vision and simulated central field loss (CFL) with eccentric viewing (EV) conditions. The solid circles are individual subject data points which have been jittered to improve visualization. The lower and upper T bars represent the minimum and maximum values and the line bisecting the boxes represent the median.

Median slope values at the x-intercept of subjects’ non-linear function under habitual vision and simulated central field loss with eccentric viewing conditions are shown in Figure 2. No signiffcant differences in the levels of reliability were found between the habitual vision and simulated central field loss with eccentric viewing conditions (Wilcoxon Signed-Rank Test, Z = −1.72, P = .09) (Figure 2). Therefore, subjects on average performed similarly irrespective of their viewing condition.

Figure 2.

Street-crossing reliability (measured as the slope at the x-intercept of subjects’ non-linear function) under normal habitual vision and simulated central field loss (CFL) with eccentric viewing (EV) conditions. The solid circles are individual subject data points which have been jittered to improve visualization. The lower and upper T bars represent the minimum and maximum values and the line bisecting the boxes represent the median.

The computed difference scores between the habitual and central field loss with eccentric viewing conditions for bias and slope were normally distributed (Shapiro-Wilk’s test of normality, W = 0.98, P = .95 and W = 0. 93, P = .16 for bias and slope, respectively). These findings are suggestive that there was no data clustering and thus the effect of the induced central field loss on bias and slope did not differ systematically from person to person.

Relationship between Street-Crossing Performance and Vision, Cognition and General Health Assessments

Significant correlations were found between bias and slope with various vision, cognitive and self-reported health measures (Table 4).

Table 4.

Spearman correlation coefficients (r) between Accuracy (measured as bias in seconds) and reliability (measured as the slope of the non-linear function at the x-intercept) with measures of vision, cognition and the eight domains of the Short Form 36 (SF-36) Health Questionnaire.

	Habitual Viewing Condition		CFL Viewing Condition
	Bias	Slope	Bias	Slope
VA (Log MAR)	−0.524*	0.312	0.233	0.410
CS (Log CS)	0.116	−0.167	Not Measured
VF Extent, radius (degrees)	0.247	−0.009	−0.010	−0.278
VF Area (degrees2)	0.198	−0.015	Not Measured
VF Remaining Area (degrees2)	0.198	−0.015	Not Measured
Scotoma Diameter (degrees)	No Scotoma		0.142	−0.111
MMSE	0.219	−0.099	−0.010	−0.298
TMT – Part B (seconds)	−0.256	−0.254	−0.095	−0.144
Physical Functioning (%)	−0.099	0.139	0.179	0.378
Role Physical (%)	−0.139	0.06	0.338	−0.020
Bodily Pain (%)	−0.171	−0.148	0.082	−0.096
General Health (%)	−0.599**	−0.111	−0.128	−0.081
Vitality (%)	−0.622**	0.009	−0.20	0.060
Social Functioning (%)	−0.210	0.042	−0.296	0.117
Role Emotional (%)	−0.162	0.214	−0.189	0.284
Mental Health (%)	−0.309	0.319	0.267	0.386

VA = Visual Acuity measured as the Logarithm of the Minimal Angle of Resolution; CS = Contrast Sensitivity measured as the Logarithm of CS; VF= Visual Field; MMSE = Mini-Mental State Exam, TMT – Part B = Trails Making Test – Part B. Vision measures were assessed monocularly under habitual normal vision and simulated central field loss (CFL) with eccentric viewing (EV) conditions.

^*indicates that correlation is significant at the 0.05 level (2-tailed) while

^**indicates that the correlation is significant at the 0.01 level (2-tailed).

Backward multiple regression analyses revealed that habitual visual acuity and the Short Form-36 Health Questionnaire domain of general health were the best predictors of bias under the habitual vision condition, accounting for 45% of the variance in bias (F(2, 17) = 8.75, P = .002, adjusted R² = 0.45). Specifically, habitual visual acuity was negatively associated with bias (b = −2.92, P = .02) suggesting that a decline in visual acuity was associated with a decrease in bias and hence less safe street-crossing decision-making. The general health domain (b = −0.02, P = .02) was also significantly associated with bias under the habitual vision condition indicating that poorer self-rated health was associated with a more positive bias. No significant predictors however, were found for bias under the simulated central field loss with eccentric viewing condition.

Backward multiple regression analyses revealed that habitual visual acuity and Trails Making Test – Part B were the best predictors of reliability under habitual vision, explaining 22.4% of the variance in slope values (F(2, 17) = 3.74, P = .04, adjusted R² = 0.22). Declines in visual acuity were associated with an increase in street-crossing reliability (b = 1.72, P = .02). Better performance on the Trails Making Test – Part B (i.e. better executive function) was associated with higher street-crossing reliability (b = −0.01, P = .04). No significant predictors were found for street-crossing reliability under the simulated central field loss with eccentric viewing condition.

Street-Crossing Survey

The street-crossing survey revealed that all subjects crossed non-signalized streets independently without assistance. Over half of the subjects (55%) reported that they crossed a moderately busy non-signalized one-way street with one lane of traffic (similar to the street assessed in this study) several times per week whereas 45% of subjects self-reported crossing this type of street only a few times per month. Ninety percent of the subjects reported no difficulty crossing a moderately busy non-signalized street similar to the type of street assessed in this study. Only 10% of subjects found it slightly difficult to cross a moderately busy non-signalized one-way street with one lane of traffic.

When self-rating their crossing behavior, 10% of subjects rated themselves as very conservative, 15% as moderately conservative, 15% as slightly conservative, 30% as slightly liberal, 25% as moderately liberal and 5% as very liberal. For perceived difficulty in judging vehicular arrival time when crossing a non-signalized street, 80% of subjects reported no difficulty whereas 15% and 5% of subjects self-reported this task as slightly and moderately difficult, respectively.

DISCUSSION

This is the first study to systematically assess the effect of an experimentally induced central scotoma on the street-crossing decision-making performance of young pedestrians in real outdoor traffic environments. We found that our young aged subjects were reliable but significantly inaccurate with their street-crossing decisions irrespective of whether or not an absolute central scotoma was present. However, under both conditions, subjects on average showed positive bias values suggesting that subjects under-estimated vehicular arrival time relative to their actual crossing times and thus they adopted a safe crossing strategy.

Finding no significant difference in accuracy or reliability between the conditions of with and without a simulated central field loss is in agreement with the findings of Hassan and Snyder,²⁷ Hassan²⁵ and Hassan and Massof²⁶ even though these earlier studies were a between-subjects design where the control group was a different group of age-matched normally-sighted subjects, the vision loss subjects were binocular and had real (i.e. not simulated) central vision loss and that their subjects were significantly older than the subjects in the current study and thus had co-morbidities and possible declines in their cognitive and/or motor skills that may have affected their performance. The agreement between the current study and these earlier central vision loss street-crossing studies is suggestive that our central field loss simulation method is an effective tool at characterizing the decision-making behavior of pedestrians with real central field loss.

While Geruschat et al.²⁴ also found that subjects with central vision loss from age-related macular degeneration were just as accurate as normally-sighted subjects in correctly identifying crossable and uncrossable opportunities, they did report that their age-related macular degeneration subjects made unsafe crossing decisions since they took significantly longer to identify crossable vehicular arrival times and had large, negative (unsafe) safety margins. A possible reason for the discrepancy in results between the findings of Geruschat et al.²⁴ and the current study may relate to the complexity of the street used in each study and the ages of the subjects. Unlike the current study, Geruschat et al.²⁴ measured crossing decisions at a large, busy, double-lane roundabout in elderly age-related macular degeneration subjects. Thus Geruschat et al.’s²⁴ findings may be explained by the fact that elderly, normally-sighted pedestrians have been shown to make significantly more unsafe crossing decisions compared to young, normally-sighted pedestrians,^41–43 most likely because of co-morbidities and declines in perceptual, cognitive and/or motor skills, and that they make significantly more unsafe crossing decisions at complex crossing environments than at simpler environments such as a one-way street.^{44, 45}

Our finding that subjects with a simulated central field loss were reliable and safe in their crossing decisions suggests that eccentric viewing can successfully compensate for any deficits in street-crossing decision-making performance as a result of a central field loss. The average visual acuity of our subjects was 20/120 and they had an absolute, monocular scotoma that ranged from 6° to 28° in diameter. Despite the presence of these large absolute scotomata and poor visual acuity, the performance of subjects when they used eccentric viewing were comparable to when they had intact, normal central vision. Numerous studies have shown that, despite the degradation of visual acuity with increasing eccentricity,^46–48 the peripheral visual field is sensitive to detecting motion and low spatial frequencies.^49–51 As the stimuli in the present study were approaching vehicles, this meant that subjects made judgements about large, moving objects that contained low spatial frequencies. Therefore, the experimental task required of subjects when using eccentric viewing with their simulated central field loss was well suited to the capabilities of the peripheral visual field. This may explain why our results did not support our hypotheses which predicted that subjects’ accuracy and reliability would worsen when subjects had a simulated central field loss.

The effectiveness of using eccentric viewing when making street-crossing decisions is further demonstrated when the results of the present study are compared to those of Wu et al²⁸ who showed that young subjects fixating centrally with a simulated, gaze-contingent central scotoma made increasingly less safe crossing decisions as the size of the central scotoma increased. Based on our findings and those of Wu et al.,²⁸ we therefore recommend that people with central field loss use a preferred retinal locus to improve the safety and accuracy of their street-crossing decisions. However, further research is required to assess the effectiveness of eccentric viewing at more complex streets since the current study assessed performance at only a relatively simple street environment.

Despite the simplistic nature of the street used in the current study, it is still important to assess street-crossing decision-making at simple crossing environments. This is because over half of the subjects in the current study reported that they crossed the simple street design assessed in this study several times every week. Additionally, the young subjects in the current study self-reported adopting a variety of street-crossing strategies ranging from “very conservative” to “very liberal”. It is therefore possible that subjects’ varying street-crossing strategies contributed to the variability in street-crossing accuracy and reliability observed in this study (see Figure 1 and Figure 2). This variation may also explain why few significant univariate correlations and predictors of performance were found between street-crossing accuracy / reliability, measures of vision, cognition and general health status. However, as found in the current study, previous street-crossing studies have also found visual acuity⁴⁴ and various cognitive function measures^{43, 52} to be significant predictors of street-crossing decision-making performance.

It is interesting to note that with the multiple regression analysis for street-crossing accuracy under habitual vision, the Short Form-36 domain of vitality did not emerge as a significant predictor of performance even though it had the highest univariate correlation with bias for the same viewing condition (see Table 4). This is most likely explained by the fact that after simultaneously adjusting for all variables in the multiple regression analysis model, vitality was no longer significant compared to the Short Form-36 domain of general health and visual acuity which had the second and third highest univariate correlations with bias under the habitual vision condition, respectively (see Table 4).

While we found that young subjects’ crossing decisions made while using their normal vision were safe, they were significantly inaccurate, a finding that is in agreement with Hassan and Snyder.²⁷ When these same subjects made decisions with a simulated scotoma, their decisions remained safe and inaccurate which is in disagreement with our previous work^25–27 which found that central vision loss subjects were on average accurate with their street-crossing decisions. A possible reason for the discrepancy in results most likely relates to the fact that the subjects in the current study were significantly younger than the central vision loss subjects in these earlier street-crossing studies^25–27 (who were at least 43 years of age and older). Certainly, Hassan and Snyder²⁷ reported that decision accuracy declined with increasing age and worsening vision. Therefore, as one ages (and there are accompanying declines in perception, cognition and motor skills) and/or as vision declines, the bias will transition from a positive value (significantly inaccurate yet safe) to a negative value (significantly inaccurate yet unsafe). Thus there will be a period, both in terms of age and level of vision loss, when the bias will be close to zero, hence resulting in it not being significantly different from zero and thus “accurate”. It is possible that the older subjects in our previous studies^25–27 were in the age that resulted in their bias being centered close to zero.

Additionally, it is very likely that the subjects in our earlier studies^25–27 had experienced their central vision loss for a long time. Consequently, these subjects might have been accustomed to their visual status and thus were able to accurately compensate for their visual deficits. This is supported by the fact that the central vision loss subjects in all of our earlier studies^25–27 self-reported that they traveled independently and crossed streets regularly without any assistance. While the subjects in the current study were also active pedestrians, they only experienced central field loss for the very first time. It is therefore possible that in response to this sudden and absolute loss of their central visual field, subjects became extra cautious with their crossing decisions which explains why more inaccurate but safer crossing performance was observed in the current study compared to those reported in our previous studies.^25–27

Limitations of the current study include the fact that subjects’ street-crossing decision-making performance was evaluated at a simple street and that the vision loss was simulated and only monocular. It is likely that subjects’ decision-making performance would have improved under binocular viewing conditions as subjects’ gap time judgments may have benefitted from the addition of binocular cues and when subjects used eccentric viewing, their ability to detect approaching vehicles would have also improved since peripheral vision improves binocularly beyond summation levels for detection and resolution tasks⁵³.

The slight increase in bias values found in the current study (1.37 sec for the habitual vision condition) compared to those obtained by Hassan and Snyder²⁷ (1.14 seconds) further support the suggestion that testing under monocular conditions resulted in subjects having slightly worse performance (i.e. greater inaccuracy). Hassan and Snyder²⁷ assessed street-crossing decision-making in a different group of similarly aged young, normally-sighted subjects at the same street using the same methods and analyses but under binocular viewing conditions. It is therefore possible that the binocular viewing conditions in Hassan and Snyder’s²⁷ study resulted in subjects being slightly more accurate. Additionally, the somewhat different manner of viewing traffic monocularly for the usually binocular subjects may have also caused them to have been more conservative with their crossing decisions, thus resulting in more positive-valued bias scores, compared to if they had used binocular vision.

Another limitation of the current study was that no mechanism was in place to monitor vehicular speed in real-time during a trial. As a result, the speed of approaching vehicles may have varied during a trial, either speeding up or slowing down. Such speed variations would alter the measured gap time and the accuracy (bias) of subjects’ crossing decisions.

To minimize drivers from altering their speed during a trial, a sign was positioned well in advance from where trials began to warn drivers about our study and that they would see people positioned by the side of the road. Additionally, experimenters monitored the speeds of all approaching vehicles both before and during a trial. Whenever an experimenter observed that the approaching vehicle’s speed changed (either during the two second sampling period or after the prompt signal), the trial was canceled and not used in any analyses.

While the above procedures minimized the risk of drivers changing speed during a trial, it is still possible that on some trials, drivers slowed down which may have contributed to subjects’ positive bias values.

Another weakness of our study was that subjects were able to visualize the simulated scotoma and hence did not experience perceptual “filling-in” as observed in patients with pathology generated central field loss. As a result, the fixation stability of our subjects when simulated with a central field loss is expected to be better than that observed in real central field loss, possibly resulting in them having better performance than subjects with real central field loss.

CONCLUSIONS

The present study showed that when subjects with an induced, absolute central scotoma adopted eccentric viewing, they were able to be as accurate and reliable in their street-crossing decisions compared to when they used their habitual, normal vision. Furthermore, all subjects, irrespective of whether or not they had an induced central field loss, adopted a safe street-crossing strategy by underestimating vehicular arrival times relative to their crossing time. These results therefore suggest that the near peripheral retina and the technique of eccentric viewing can be used successfully to make appropriate street-crossing decisions.

We found that visual acuity, and self-rated general health were significant predictors of street-crossing accuracy while visual acuity and a cognitive measure were predictive of street-crossing reliability. Finding visual acuity and a cognitive measure to be predictive of street-crossing performance is in agreement with earlier street-crossing studies.^43,44,52