Impact of macular scotoma and tubular vision on oculomotor behavior and performance in visuospatial comparison tasks

Abstract

Our aim in this study was to understand how we perform visuospatial comparison tasks by analyzing ocular behavior and to examine how restrictions in macular or peripheral vision disturb ocular behavior and task performance. Two groups of 18 healthy participants with normal or corrected visual acuity performed visuospatial comparison tasks (computerized version of the elementary visuospatial perception [EVSP] test) (Pisella et al., 2013) with a gaze-contingent mask simulating either tubular vision (first group) or macular scotoma (second group). After these simulations of pathological conditions, all participants also performed the EVSP test in full view, enabling direct comparison of their oculomotor behavior and performance. In terms of oculomotor behavior, compared with the full view condition, alternation saccades between the two objects to compare were less numerous in the absence of peripheral vision, whereas the number of within-object exploration saccades decreased in the absence of macular vision. The absence of peripheral vision did not affect accuracy except for midline judgments, but the absence of central vision impaired accuracy across all visuospatial subtests. Besides confirming the crucial role of the macula for visuospatial comparison tasks, these experiments provided important insights into how sensory disorder modifies oculomotor behavior with or without consequences on performance accuracy.

Introduction

Cultural evolution was accompanied by the invention of graphic symbols providing semantic, quantitative, schematic, or directional information. These visual signs, used to write, count, or draw, have been designed to be highly informative while easily made with wedges on tablets or ink on paper (Daniels, 1996) and so tend to be peculiar visual objects with various defined spatial combinations of strokes, sometimes completed with dots. The use and understanding of such complex marking systems depends not only on specific explicit learning (Dehaene & Cohen, 2007) but also on visuospatial discrimination. Indeed, differences in length, size, orientation, and relative position of separable features may be highly meaningful. 

The elementary visuospatial perception (EVSP) test (Pisella et al., 2013) (Figure 1, left column) has been developed to evaluate fine visuospatial discrimination of pairs of symbols across a series of subtests. It consists of comparing, through vision, the horizontal extent of two black geometrical shapes made up of two straight lines (subtest T1), the size of two black discs (subtest T2), and the relative orientation (subtest T3 of angle comparison) or position in space (subtests T4, T5, and T6) of the visual components of the symbols. Even though these subtests appear simple, children's performance across various ages has shown that success on these tasks depends on the neurodevelopmental maturity of the brain. Adult-level performance was reached only by children between 6 and 8 years old for length and size comparisons, at 12 years of age for dot localization, and even later for the ability to judge angles and midlines, suggesting further maturation through adolescence (Pisella et al., 2013). The EVSP has been validated as a screening test for children with neurodevelopmental disorders where deficits in visuospatial perception may play a contributory role (Pisella, Martel, Roy, Vuillerot, & Gonzalez-Monge, 2020; Pisella et al., 2021) and for adults with occipitoparietal lesions from stroke presenting with Balint syndrome (Pisella et al., 2013). Accordingly, neuroimaging studies have shown that such apparently simple visual judgments about object metrics and spatial relationships require the involvement of high-level areas of the dorsal visual stream (e.g., Faillenot, Sunaert, Van Hecke, & Orban, 2001; Faillenot, Toni, Decety, Grégoire, & Jeannerod, 1997; Zachariou, Klatzky, & Behrmann, 2014). Their necessary contribution has further been established by the observation of impaired accuracy of perceptual judgments after occipitoparietal damage (Fink et al., 2000; Harvey, Milner, & Roberts, 1995; Kerkhoff & Marquardt, 1995; Pisella et al., 2013; Ungerleider & Mishkin, 1982). 

Figure 1.

 

Illustration of the EVSP test (with an example for each subtest) performed in full vision (left column), artificial tubular vision (middle column), and artificial macular scotoma (right column) conditions. Left column: For T1, objects consisted of horizontal lines of various extents, from 0.7 to 4.5 cm, combined with a vertical bar of 1 cm placed either at the left extremity of the horizontal line (forming a L shape) or at 1.5 cm from the left extremity of the horizontal line (forming an asymmetrical T shape upside down, as illustrated here). The closest distance between the two objects aligned horizontally was 0.9 cm. Participants had to judge whether or not the two objects had the same length. For T2, objects consisted of two discs of the same or different diameter, from 0.75 to 2.25 cm. The closest distance between the two objects aligned horizontally was 1.8 cm. Participants had to judge whether or not the two discs were the same size. For T3, objects consisted of two crosses; the right cross was composed of two perpendicular bars of 4 cm, one horizontal and one vertical, and the left cross displayed either a near-horizontal or a near-vertical bar leading to angles that could be 90° ± 7°. The closest distance between the two objects, not aligned horizontally, was 2 cm. Participants had to judge whether or not the two objects had similar angles. For T5, objects consisted of two square frames of 6 cm with a black dot of 3-mm diameter inside each frame with a positional difference ranging from 3 to 6 mm. The closest distance between the two frames, not aligned horizontally, was 7.5 cm. Participants had to judge whether the dot was at the same relative position to the frame in the two objects. Subtest T6 consisted of the same right object as T5, but the left one contained several randomly arranged colored dots. Participants had to select the dot corresponding to the position of the black dot in the right frame and report its color before pressing any key to initiate the next trial. Only one example, with three colored dots, was presented before starting subtest T6. The task became more complex as the number and proximity of the dots in the left square increased (from three dots in trial A to 10 dots in trial L). The midline judgment subtest (T4) consisted of a single pre-bisected horizontal line of 12 cm (16.7°) with spatial offsets from the true midline ranging from 0 to 6 mm. The examples presented beforehand consisted of one with a bisection mark clearly not at the middle (at 3 cm from the true midline) and another one with the bisection mark at the true midline. Participants were asked to press the letter “P” if they judged the bisection mark to be at midline and the letter “Q” if they judged that it was not. Middle column: Simulated tubular vision of 10° corresponded to a gaze-contingent visibility window of 7.18-cm diameter. For all subtests involving two distinct objects (T1, T2, T3, T5, and T6), this gaze-contingent mask hid the one object when fixating the center of the other object, except for four among the 12 trials of subtest T1 and one among the 12 trials of subtest T2. For subtest T4, this gaze-contingent mask hid the two extremities of the horizontal line of 12 cm (16.7°) when fixating the bisection mark in each trial. Right column: Simulated macular scotoma of 10° corresponded to a gaze-contingent central circular mask 7.18 cm in diameter. For subtests T1 and T2 involving two distinct objects in close proximity, this artificial macular scotoma hid both objects, at least partially when the participant attempted to explore them using central vision, forcing them to use peripheral vision through an eccentric fixation. For subtests T3, T5, and T6, involving two distinct objects of 8.04° for T3, and 11.80° for T5 and T6, when the participant fixated one object (hidden by the macular scotoma), the gaze-contingent mask did not hide the other. For subtest T4, when the participant fixated the center of the horizontal line of 12 cm (16.7°), the gaze-contingent mask hid the bisection mark but allowed the participant to see the two extremities of the line.

How does the human visual system deal with the challenge of the spatial discrimination of two written symbols? To process visual information, one can move the eyes to each relevant stimulus or each relevant component of a stimulus in order to process them serially using central vision and/or use simultaneous static information from both peripheral and central vision. There is a functional and anatomical difference between central and peripheral vision from the retinal to the cortical visual processing level (Stewart, Valsecchi, & Schütz, 2020). The macula is a retinal zone measuring about 5.5 mm in diameter that is responsible for central vision. This small region, extending to about 10° of visual eccentricity from the fixation point, contains half the ganglion cells of the entire retina with a high density of cones, which gives it high visual acuity (Coletta & Williams, 1987; Duke-Elder & Shimkin, 1926) and a magnified representation within the primary visual cortex. The rest of the retina, corresponding to the outer 20° of visual eccentricity, contains the other half of the ganglion cells, with a small density of cones, and thus has lower visual acuity (Coletta & Williams, 1987), as well as a much smaller representation in the primary visual cortex. Although the former is important for fine perceptual discrimination, the latter is preferred for obtaining global image characteristics and guiding saccades to distant informative locations (Rayner & Bertera, 1979; Rayner & McConkie, 1976; Van Diepen, Wampers, & d'Ydewalle, 1998). Peripheral and central vision are complementary and interact with each other (Stewart et al., 2020), and for object search in naturalistic visual scenes, neither macular nor peripheral vision has proven to be crucial (Nuthmann, 2014). But, the inhomogeneity of the visual system presents a significant potential bias for fine visuospatial discrimination. The magnitude and shape of the visual input representation in primary cortical areas differ dramatically depending on whether it is viewed with the macula or the retinal periphery (Aagten-Murphy & Bays, 2018; Valsecchi & Gegenfurtner, 2016). Positioning the eyes between two objects to be compared, such that both are viewed at an equivalent visual eccentricity, and therefore experiencing an equivalent cortical magnification distortion appears to be an optimal strategy. Healthy participants can perform size or length comparisons accurately when required to keep their gaze fixated either in-between two objects presented at the same eccentricity in the left and right visual fields (Cappelletti, Freeman, & Cipolotti, 2009) or on only one of the two objects (Jurkiewicz, Salemme, Froment, & Pisella, 2021). However, this is not the strategy they spontaneously adopt in free-gaze conditions; instead, they exhibit a gaze exploratory strategy with oculomotor routines that do not necessarily enhance the accuracy of their perceptual judgments (Aizenman, Gegenfurtner, & Goettker, 2024). Active scanning involves saccades alternating between the objects to compare, suggesting a proactive sequential delivery of objects to the macula for high-acuity visual processing. But, when one object is fixated, the other one can only be compared from peripheral vision or from the holding of information of the object not currently fixated in visual memory. Therefore, transsaccadic association in visual memory is required to ensure spatial consistency across space (Herwig & Schneider, 2014), between macular and peripheral input, and between sequentially sampled macular information. However, a high frequency of comparative saccades has also been suggested to reduce memory usage (Galpin & Underwood, 2005). This raises the second issue we wish to address in the present paper, which is the functional compensation of visual field defects in visuospatial comparative tasks. Whereas partial visual masking increases the cost of information acquisition and memorization, study participants have often remained able to perform with comparable accuracy but with a general increase of the time needed to complete one trial (Hardiess & Mallot 2015; Hardiess, Papageorgiou, Schiefer, & Mallot, 2010; Janssen & Verghese, 2015). This cost in response time can emerge from different spontaneous or deliberate oculomotor compensatory strategies with respect to the task-specific demands and type of visual loss. 

Here, we designed a computerized version of the EVSP screening test with a two-alternative forced-choice button press responses (same/different), eye tracking, and gaze-contingent masking. This allowed us to record spontaneous eye movements performed during the EVSP test and investigate whether accuracy at each subtest crucially depends on peripheral vision, macular vision, or both. The aim was to understand the high-level oculomotor and visual processes involved in visuospatial judgments and to provide insights into how patients with retinal disease affecting their peripheral versus macular vision (e.g., retinitis pigmentosa and Stargardt disease, respectively) may undertake visuospatial comparisons. 

Methods

Participants

Eighteen healthy participants gave their informed consent for the first experiment (mean age, 26.06 years; SD = 4.79; 60% females and 40% males) and another 18 participants for the second experiment (mean age, 27.58 years; SD = 2.33; 60% females and 40% males). All had normal or corrected visual acuity and no neurological damage or visual pathologies. This study protocol was promoted by the University Hospital of Lyon and received the approval of the Northwestern 1 046/2017 Research Ethic Boards (registration no. 2017-A02562-51). 

Experimental setup

Participants performed the experiments in a semi-dark room, seated with their eyes at a distance of 40 cm from the computer screen (Dell P Series 21.5-inch flat panel monitor, P2219H; 47.6 × 26.8 cm; refresh rate, 60 Hz; resolution, 1920 × 1080; Dell, Round Rock, TX) and their forehead and chin stabilized. The right eye position was recorded using a high-speed eye tracker (EyeLink Portable Duo Eye Tracker, 1000 Hz; SR Research, Ottawa, Canada). The experiments were developed, displayed, and recorded (button press as well as eye movements) with Experiment Builder software (SR Research). Before each test for each viewing condition, a five-dot calibration scaled to the screen was performed. 

Stimuli and tasks

The EVSP test was comprised of six subtests, T1 to T6, illustrated by an example in Figure 1 (left column) and described in detail in the legend for Figure 1. Each subtest consisted of 12 trials of increasing difficulty. Participants could take as much time as they liked. For each trial, the stimuli remained visible until the participant made a response consisting of a key press: the letter “P” on the keyboard if they judged the two objects to be the same (or the mark being at the true middle in T4) and “Q” if they judged them to be different (or not at midline in T4). For T6, participants had to verbally report the color of the dot corresponding to the single black dot in the right frame, and press any key. 

Because we aimed at simulating a clinical condition in which individuals never had the opportunity to see the test objects in full vision, all participants undertook task with all subsets first in the gaze-contingent masking condition without any feedback about their performance. In Experiment 1, participants performed the EVSP test with a gaze-contingent peripheral mask (see Figure 1, middle column) which simulated tubular vision (i.e., a circular central visual field limited to a diameter of 10° of eccentricity and always centered on the gaze position). In Experiment 2, participants performed the EVSP test with a gaze-contingent central mask with a 10° diameter (see Figure 1, right column) which simulated a macular scotoma. Then, after a short break, they undertook the task a second time but in full vision. This order of tasks should favor the full vision condition, highlighting the strength of results showing no deficit in the simulated “pathological” condition compared with full vision. 

Spatial comparison performance parameters

Based on the button press of participants in each trial, two parameters were extracted: 

• • A performance score was computed over the 12 trials of each subtest. For each trial, a score of 0 was assigned when the response was false or 1 when it was correct. This led to a performance score out of 12. For all subtests except T6, the response was two-alternative forced-choice (same/different) with a chance level at 6/12. For subtest T6, a verbal response was collected manually by the experimenter corresponding to a selection among the proposed colored dots in each trial (between three and 10), so chance level was different (between 4 and 1.2) depending on the number of colored dots.
• • A mean response time in seconds was computed over the 12 trials of each subtest.

Saccade analysis

Eye movements were visualized using the spatial overlay view of the DataViewer (SR Research) software, which traces, on the stimulus displayed on the screen, the saccades as yellow lines and the fixations as blue circles, with diameter reflecting their duration (examples are shown in Figure 2). DataViewer software also provides saccade parameters for offline analysis, and saccade detection was set at an acceleration threshold of 4000°/s/s. For each trial of each subtest, we analyzed the following oculomotor parameters: the number of saccades, their amplitudes, and latencies. Aberrant saccades were removed manually based on visual inspection of the ocular traces (very large saccades associated with blinking, saccades performed off-screen) before further data processing. 

Figure 2.

 

Typical examples of eye movements recorded for each EVSP subtest in the full vision condition. Each example displays the ocular trace of a different participant on the same trial reflecting the variety of behavior observed for each subtest. Blue circles correspond to fixation positions (diameter reflecting the duration of fixation) and yellow lines correspond to saccadic movements sampled at 1000 Hz. Subtest T1: Exploration of the lines never reached the extremities and often took place only on one object as if it was taken as reference; ES only occurred on one object, and the other object tended to be sampled with AS fixating it many times at a preferred location, often at the intersection of the two lines constituting the object or at the position equivalent to where the exploration of the reference object horizontal line ended. Both types of saccades were mainly horizontal (right/left) with small vertical variation, corresponding to the context: The two objects were horizontally aligned and they could differ only on the length of the horizontal line. Subtest T2: Typically, AS between the two discs seemed to aim the center or the contour of the discs with little variation. Unlike T1, participants explored the two objects more equally with small-amplitudes ES. Subtest T3: To compare the orientations of the lines constituting the two crosses, AS were diagonal and of large amplitude. The participants’ typical exploration strategy seemed to be to aim their gaze at the intersection of the two lines constituting each cross to compare the angles. However, some participants also fixated a location where the tilt of the line would appear more pronounced than at the intersection, or performed ES along one line in order to discern its orientation. Subtest T4: For this midline judgment subtest involving a long horizontal line and a small vertical bisection mark, saccades were of short amplitude (3.65 on average, similar to T1 and T2) relative to the length of the line. For the majority of the participants, they seemed to correspond to a progressive exploration of the horizontal extent on each side from the bisection mark. Exploration of the line was incomplete; it never reached the two extremities of the line. Fixations tended to be centered symmetrically around the mark in about one-third of the participants, but was more frequently asymmetrical (i.e., more concentrated on one side). Subtest T5: For this subtest requiring the comparison of two objects made up of a dot within a square frame, participants performed numerous saccades, mostly large AS (diagonal alternating between the two objects in a spatial configuration similar to T3), mainly aiming the two dots, and small additional saccades exploring the distances between the dots and the edges of the frames. Subset T6: For this subtest involving comparison of the position of the dot in the right frame and the positions of multiple colored dots in the left frame, the greater number of saccades performed may reflect the more numerous spatial comparisons involved in this subtest, compared to T5. Indeed, in order to be able to select, among the multiple colored dots provided, the one lying at the exact corresponding location of the black dot in the right frame, several dot–frame relative position comparisons are required either within the left frame (ES) or between the left and the right frame (AS). The oculomotor strategy is similar to that of subtest T5, mainly involving precise large-amplitude saccades repetitively alternating between the right single dot and the most probable corresponding left dots, and smaller saccades exploring the relative distances between the dots and the edges of the frames.

Saccade categorization

For the subtests involving the comparison of two objects, two categories of saccades were identified: exploration saccades (ES), defined as small amplitude saccades performed within an object, and alternation saccades (AS) of larger amplitude performed between objects. This categorization was adjusted for each subtest based on object size. For subtests T1 and T2 in which object size varied, the categorization was even adjusted for each trial. In order to compare the distribution of saccade amplitudes for the entire subtests T1 and T2 between full vision and artificially masked vision conditions, object sizes were normalized to 0°, allowing us to merge the amplitude data for the 12 trials. Saccades of negative amplitude therefore correspond to saccades smaller than the extent of the object (i.e., ES), whereas saccades of positive amplitude correspond to saccades larger than this limit (i.e., AS). For T4, only ES along the single horizontal line of 12 cm were identified. In Experiment 2, a third type of saccade was identified—saccades with very large amplitudes allowing eccentric fixations (EFS) of the object. This additional category of saccades was defined by amplitudes greater than the radius of the artificial macular scotoma (5°), allowing the participants to shift their gaze (and artificial scotoma) away from the object and to process it from the nearest eccentric fixation using peripheral vision. 

Statistics

We performed t-tests to compare EVSP performance scores and response times for each subtest (T1, T2, T3, T4, T5, and T6) between full vision and artificial tubular vision conditions in Experiment 1 and between full vision and artificial macular scotoma conditions in Experiment 2. Alpha was set at 5%, and Bonferroni correction for 12 multiple comparisons (six subtests and two parameters: EVSP score and response time) was applied for Experiment 1 and Experiment 2 separately, setting the statistical threshold at alpha/12 = 0.00416. Effect sizes (Hedge's g) and equivalence Bayesian testing (Supplementary Tables) are also provided for these paired-sample t-tests. The change of the distribution of the different saccade categories between viewing conditions was tested using Pearson’s χ² tests comparing the ratio of ES produced by the group of subjects among their total number of saccades between full vision and tubular vision in Experiment 1 and between full vision and macular scotoma in Experiment 2. These χ² tests were performed on the website biostatgv.sentiweb.fr. 

Results

Figure 2 illustrates, for each EVSP subtest, two typical examples of eye movements recorded for different participants for the same trial in the normal viewing condition and reflects the small variety of observed behavior. The average number of saccades by trial is provided for each subtest in Table 1 for full vision and also for artificial pathological conditions. As can be seen qualitatively in Figure 2 and quantitatively in Table 1, the apparently simple spatial judgments of the stimuli pairs of the EVSP screening test involved a minimum of six saccades per trial in full vision. The relative distributions of ES and AS and their average amplitudes are provided for each subtest and each viewing condition in Tables 1 and 2, respectively. 

Table 1.

 

Average total saccade numbers (n) and average proportion of saccades for each category (%ES, %AS, %EFS) by condition and subtest.

 

Table 2.

 

Average saccade amplitude (°) within each category for each condition and subtest.

 

Experiment 1: Comparison between artificial tubular vision and full vision

Figure 3 shows the effect of gaze-contingent peripheral vision masking on response accuracy and time. The absence of peripheral vision only affected accuracy in subtest T4, t(17) = −4.59, p = 0.000259, Hedges's g = 1.48. Although the mean score across participants was 9.39/12 of correct judgment in full vision, it was close to chance level (7.5/12 on average) with the artificial tubular vision. For all other subtests, there were no significant differences between accuracy scores in the tubular vision and full vision conditions, with all accuracy scores > 9.5 on average, all t(17) > −1.56, all p > 0.13, and all Hedges's g < 0.4. 

Figure 3.

 

Boxplots of accuracy and response time for each EVSP subtest in Experiment 1. The top panel shows the mean accuracy, standard error (box), and confidence interval (whiskers) for full vision and tubular vision. The bottom panel shows the mean response time, standard error (box), and confidence interval (whiskers) for full vision and tubular vision.

Tubular vision significantly increased the mean response times compared to full vision (Figure 3, bottom panel) for subtest T3 angle comparison, t(17) = 5.22, p = 0.000007, Hedges's g = 1.38; T4 midline judgment, t(17) = 5.53, p = 0.000036, Hedges's g = 1.37; T5 relative position comparison, t(17) = 5.05, p = 0.000097, Hedges's g = 1.07; and T6 position selection, t(17) = 5.69, p = 0.000026, Hedges's g = 1.29. In contrast, there were no significant differences between tubular vision and full vision in mean response times for the subtests of length and size comparison: T1, t(17) = 3.29, p = 0.00429, Hedges's g = 0.81; T2, t(17) = 2.38, p = 0.04479, Hedges's g = 0.51. 

In order to analyze the ocular behavior of the participants in terms of small-amplitude ES and large-amplitude AS, the saccade amplitude distribution of the group was plotted for full vision and tubular vision. As can be seen in Figure 4, this comparison revealed a decrease of AS and an increase of ES in the absence of peripheral vision. This is particularly clear for subtests T3, T5, and T6, which displayed a bimodal distribution of saccade amplitudes, the smaller corresponding to ES and the larger to AS. In tubular vision, participants produced about three-quarters fewer AS and almost doubled their ES. 

Figure 4.

 

Saccade amplitude group distribution in full vision (orange bars) compared with tubular vision (blue bars) for each EVSP subtest. ES are defined as having an amplitude smaller than the object, whereas AS performed between one object and the other were of greater amplitude because they alternated roughly between the object centers. The object size was fixed in T3, T5, and T6 and was variable among trials for T1 and T2. In order to compare the distribution of saccade amplitudes for the entire subtests T1 and T2 between full vision and artificially masked vision conditions, these limits were normalized to 0° for each trial, allowing us to merge the amplitude data for the 12 trials. Saccades of negative amplitude therefore correspond to saccades smaller than the extent of the object (i.e., ES), and saccades of positive amplitude correspond to saccades larger than this limit (i.e., AS).

For subtest T4, which did not involve two distinct objects but two sides of a single line to compare so there were no AS, tubular vision globally increased the number of saccades, and the larger amplitude ones (>7°) almost disappeared. Even though subtests T1 and T2 involved two distinct objects, the distribution of saccade amplitudes was not bimodal because the distance between the two objects was too similar to their magnitude, which moreover was variable. The normalization of saccade amplitude by object magnitude allowed us to separate AS and ES, though, by assigning them positive and negative amplitudes, respectively. A slight oculomotor behavior modification could be observed with more numerous ES and fewer AS in tubular vision. However, this appears more as a shift of the distribution and a global increase of the number of saccades for these subtests. The change in the ratio of the total number of ES among the total number of saccades at the level of the group of subjects of Experiment 1 between full vision and tubular vision conditions (illustrated in Figure 4) was significant for all subtests with Pearson's χ² tests (all p < 0.002). 

Experiment 2: Comparison between artificial macular scotoma and full vision

Effect of gaze-contingent macular vision masking on response accuracy and time

As seen in Figure 5, there was a significant decrease in accuracy with the absence of macular vision for the T1 length comparison, t(17) = −3.33, p = 0.003948, Hedges's g = 0.72; T2 size comparison, t(17) = −3.85, p = 0.00127, Hedges's g = 1.03; and T4 midline judgment, t(17) = −4.97, p = 0.000116, Hedges's g = 1.63. In contrast, there was no significant decrease in accuracy in macular scotoma compared to full vision for the subtest T3 angle comparison, t(17) = −2.55, p = 0.020401, Hedges's g = 0.54; T5 relative position comparison, t(17) = −1.84, p = 0.082698, Hedges's g = 0.55; or T6 position selection, t(17) = −3.28, p = 0.004331, Hedges's g = 0.7. As can be seen in Figure 5 (bottom panel), the macular scotoma significantly increased response times compared to full vision in all subtests, with all t(17) > −3.98, all p < 0.000308, and Hedges's g > 0.96. 

Figure 5.

 

Boxplots of accuracy and response time for each EVSP subtest in Experiment 2. The top panel shows the mean accuracy, standard error (box), and confidence interval (whiskers) for full vision and macular scotoma. The bottom panel shows the mean response time, standard error (box), and confidence interval (whiskers) for full vision and macular scotoma.

Effect of gaze-contingent macular vision masking on oculomotor behavior

As can be seen in Figure 6, the comparison of the distribution of saccade amplitudes in macular scotoma compared to full vision revealed an increased number of the larger amplitude saccades (in all subsets) in combination with either an absence (in T1, T2, and T4) or a decrease (in T3, T5 and T6) of the number of smaller amplitude saccades. For subtests T1, T2, and T4, the main observation was the appearance of a long tail of large-amplitude saccades (>5°) made by participants in order to have an eccentric fixation compensating for the artificial macular scotoma (EF). For subtests T3, T5, and T6, which showed a bimodal distribution of saccade amplitudes, results appeared to be opposite those of Experiment 1, as participants approximately doubled their number of AS, and there was a visible decrease in the number of ES with gaze-contingent macular vision masking. The changes in the ratios of the total number of ES among the total number of saccades at the level of the group of subjects of Experiment 2 between full vision and macular scotoma conditions (illustrated in Figure 6) were significant for all subtests (not tested for T4, where only ES are produced in full vision) with Pearson's χ² tests (all p < 2 × 10⁻⁷). 

Figure 6.

 

Same as Figure 4 except that the blue bars correspond to the macular scotoma condition, and eccentric fixation saccades (EF) appeared in subtests T1, T2, and T4 characterized by amplitudes larger than 5°, the value corresponding to the radius of the circular artificial macular scotoma.

Discussion

Two groups of healthy participants performed the computerized EVSP test with gaze-contingent masking of peripheral or central vision in Experiment 1 (tubular vision) and Experiment 2 (macular scotoma), respectively. Both groups subsequently performed the EVSP test in full vision. Performance was compared between full vision and masked conditions in terms of response accuracy, response time, and ocular behavior to obtain insight into the respective roles of macular and peripheral vision in perceptual tasks involving visuospatial comparisons. The findings also provide insight into how patients with retinal lesions affecting central or peripheral vision may deal with visuospatial comparison tasks. 

The extent to which macular and peripheral vision contributed significantly to visuospatial ability varied depending on the subtests. Performance for the subtests of length and size comparison (T1 and T2) relied mainly on macular vision, as participants were able to perform them without a significant change in response accuracy and time despite an artificial tubular vision of 10°. The subtests of spatial relationship comparison (T3, T5, and T6) could still be completed using either exclusively central or peripheral vision without a significant decrease in accuracy, albeit with a significant increase in response time. The same preservation of accuracy with an artificial blindspot (macular scotoma) or spotlight (tubular vision) of similar size has also been observed for visual searches performed on real-word visual scenes (Nuthmann, 2014). Only the ability to judge whether a long line of 12 cm was correctly pre-bisected (subtest of midline judgment T4) required both central and peripheral vision; there was a drastic decrease in accuracy when one or the other was artificially masked compared to full vision. T4 was the only subtest involving a single object (the line) rather than two but nevertheless required a comparison between the two sides (on the left and on the right of the pre-bisection mark). Participants typically achieved it in full vision by progressively exploring the horizontal extent of the line from the bisection mark on one side or both, as revealed by small incremental ES that never reached the two extremities of the line (Figure 2). Accuracy during T4 was not maintained in the absence of peripheral vision, despite an increase in the number of ES aimed at examining the spatial extent of the line up to its extremities on both sides from the bisection mark (Figure 3). Accuracy also dropped in the absence of macular vision (Figure 5) despite the strategy of eccentric fixation above or below the pre-bisection mark in order to be able to process the entire line in peripheral vision. Taken together, the results of Experiments 1 and 2 showed that it was almost impossible for the participants to succeed in the midline judgment task without being able to attend covertly to the line extremities while fixating the pre-bisection mark, thus requiring both central and peripheral vision. 

All other subtests involved the comparison of two separate objects and were performed with small-amplitude ES within objects and large-amplitude AS between the two objects. The main result of Experiments 1 and 2 altogether (Figures 4 and 6) was that the number of ES depended on macular vision, whereas the number of AS depended on peripheral vision. Indeed, in the condition of artificial tubular vision of 10° (Experiment 1), the pool of the largest AS decreased (Figure 4). More precisely, for subtests T1 and T2 there was no drastic modification of oculomotor behavior, response accuracy, or time with the artificial tubular vision of 10°. The AS remained mostly unchanged for these subtests because, when an object was fixated, the other one could still be processed in macular vision thanks to the proximity and the small size of the two objects. In contrast, in subtests T3, T5, and T6, where the two objects to be compared were larger and farther away, the large AS almost disappeared with the artificial tubular vision of 10°. Conversely, in the artificial macular scotoma condition (Experiment 2), AS were more numerous but the number of ES decreased. Increasing AS allowed compensating for the macular scotoma in subtests T3, T5, and T6, where participants maintained good response accuracy, but not in subtests T1, T2, and T4. In T1 and T2, the artificial macular scotoma hid both objects as soon as one was fixated upon in most trials, leading to the production of EF in order to simultaneously process the two objects in peripheral vision. EF were characterized by an amplitude larger than 5°, a value corresponding to the radius of the artificial macular scotoma. They appeared in T1, T2, and T4 as the only way to perform the task but did not allow participants to achieve a response accuracy similar to that of full vision. 

In T3, T5, and T6, where the objects were farther apart, participants managed to maintain accuracy similar to that for full vision, but response times significantly increased. We suggest that maintaining response accuracy despite the macular scotoma in T3, T5, and T6 through an increase in the number of AS may be enabled by presaccadic perceptual facilitation, a phenomenon that offers an optimized visual processing in peripheral vision just before saccade execution (Deubel & Schneider, 1996). This phenomenon appears as a particularly efficient covert shift of attention toward saccadic goal location (Khan, Blohm, Pisella, & Munoz, 2015; Ouerfelli-Ethier, Comtois Bona, Fournet, Pisella, & Khan, 2023), potentially providing better perception than simply fixating and covertly attending to the peripheral stimulus. In terms of the underlying neural process, it is sometimes attributed to an anticipatory shift of macular vision receptive fields toward the future gaze position before the eyes have moved, as revealed by monkey electrophysiology in a region of the visual dorsal stream (presaccadic remapping) (Duhamel, Colby, & Goldberg, 1992). There is some evidence in humans that covert attention and presaccadic remapping are dissociated neural processes (Blangero et al., 2010) and that presaccadic perceptual facilitation operates at a retinal level of processing related to sensory acuity (Smith, Beierholm, & Avery, 2023), as significant perceptual impairments have been reported at the postsaccadic location of the blind spot before saccade onset. It could thus be conversely argued that, in case of non-retinal (artificial) central scotoma, a saccade aimed directly toward a peripheral location could still offer, just before saccade execution, processing with good central vision acuity. This would explain the maintenance of response accuracy in the macular scotoma condition for subtests T3, T5, and T6, where the other object is seen via peripheral vision during the AS. AS would therefore have been spontaneously used in the condition of macular scotoma to process the other object from eccentric fixation while fixating one object. The fact that response time, however, increased compared to full vision can be interpreted as follows: AS performed using both macular and peripheral vision allow for a direct comparison between objects, one in central and the other in peripheral vision, during the period of time preceding the AS. It can thus be argued that in full vision, just before their execution, AS allow both objects (the one directly gazed and the postsaccadic one) to benefit simultaneously from the highest central vision acuity. This would explain why AS were also produced in subtests T1 and T2, even though both objects could be apprehended simultaneously in macular vision by simply fixating in-between them, which could have been a good strategy allowing the processing of both objects at low and comparable visual eccentricity. 

This particular aspect drew our attention. Why did participants not simply take a single simultaneous snapshot of the two objects in only one visual fixation at a privileged location when the objects were at a short visual distance from the fovea in the visual field? Indeed, it has been shown that healthy participants can accomplish size or length comparisons while maintaining gaze fixation, when this is required in an experiment, either in-between two objects presented at the same eccentricity in left and right visual fields (Cappelletti et al., 2009) or on one of the two objects (Jurkiewicz et al., 2021) with a small size eccentricity underestimation bias. However, this is clearly not the spontaneous behavior that we observed in the free-gaze condition. Participants, instead, produced a large number of ES and AS in each trial and each subtest. This active vision, in which the eyes rarely remain still during visual perception, has been repeatedly revealed by the oculomotor recordings since the work of Yarbus (1967). However, why this active vision is beneficial compared with static vision remains unclear. We wish to suggest that the dorsal visual stream plays a crucial role in this active vision, explaining some of its benefit, especially for visuospatial perception, by (1) providing an efficient spatiotemporal integration of snapshots across saccades (Cheviet, Pisella, & Pelisson, 2021; Colby, Duhamel, & Goldberg, 1995; Colby & Goldberg, 1999; Deubel, Schneider, & Bridgeman, 2002; Pisella, 2017; Pisella & Mattingley, 2004; Pisella et al., 2011) and (2) implementing an active “peripheral magnification” in its spatial representation to match the real metrics of the visual environment for accurate action (Vindras et al., 2016) and perception (Jurkiewicz et al., 2021). First, we have speculated that presaccadic remapping processes, which depend on the visual dorsal stream (Blangero et al., 2010; Duhamel et al., 1992), allow the comparison of central and peripheral objects with the same central vision acuity. In favor of this interpretation, it has been shown that saccade generation can compensate for the size-eccentricity effect (Laaboudi, Hillairet De Boisferon, & Paeye, 2023; for other studies showing calibration of shape perception prior saccade, see Herwig, Weiß, & Schneider, 2015; Paeye, Collins, Cavanagh, & Herwig, 2018). This perceptual bias of objects in the periphery being perceived smaller than objects in central vision (Baldwin, Burleigh, Pepperell, & Ruta, 2016; Kirsch, Pfister, & Kunde, 2020) is attributed to the greater number of photoreceptors in the macula compared with the peripheral retina, causing an overrepresentation of central vision in the visual cortex (also referred to as central cortical magnification) (Daniel & Whitteridge, 1961), along with peripheral vision compression in cortical space representation. However, saccade preparation could overcome this perceptual magnitude distortion in the visual field. Indeed, the posterior parietal cortex has been described as displaying representations of the visual space without a magnified representation of macular vision relative to peripheral vision (Fattori et al., 2009; Galletti, Fattori, Gamberini, & Kutz, 1999; Pitzalis, Fattori, & Galletti, 2015). These accurate spatial representations would emerge from the transsaccadic remapping processes and would be constantly updated using visuospatial working memory during active vision (Aagten-Murphy & Bays, 2018; Colby et al., 1995; Khan, Pisella, Rossetti, Vighetto, & Crawford, 2005a; Khan et al., 2005b; Pisella, 2017). Based on these combined processes of active compensation of perceptual distortions in the visual field and visuospatial synthesis of successive snapshots in dynamic spatial representations, active vision would help to refine spatial perception. 

With regard to the potential clinical interest of this study, interesting information can be gained by transposing the results of Experiment 1 to an impairment of the peripheral visual field (e.g., retinitis pigmentosa) and the results of Experiment 2 to a macular scotoma (e.g., Stargardt disease). In this context, the speculations proposed here about the role of oculomotor behavior in overcoming the distortions due to central magnification and building global spatial representations from local sampling of visual information may guide future rehabilitation strategies and enrich clinical practice. For example, the benefit of making AS was drastically reduced in the case of artificial tubular vision, because the absence of peripheral vision hindered the ability to accurately target the peripheral object. This highlights the interest in clinically training patients with retinitis pigmentosa to elicit such saccades outside their residual central visual field, as centrifugal saccades allow the perception of a larger visual scene with central acuity resources applied to peripheral objects not only after but also just before their execution. They should also benefit from intact dorsal visual stream processes of spatial synthesis of these large-scale multiple visual snapshots. Conversely, AS increased in the case of artificial central scotoma because they allowed the participant to overcome the absence of central vision thanks to presaccadic perceptual remapping, which utilized central acuity resources at least prior to onset. This finding provides important insights into how presaccadic perceptual remapping may help to compensate for some neurovisual deficits due to cortical damage, such as hemianopia, and not due to retinal disease. If presaccadic perceptual remapping relies on a cortical organization reflecting the photoreceptors of the retina, it may still be present and beneficial in cases of retinal diseases affecting central vision after a normal neurological development such as age-related macular degeneration or Stargardt disease. It should, however, be mentioned that gaze-contingent masking simulates the clinical conditions only in terms of how retinal loss affects the sampling of visual information and not in terms of how its chronicity can induce behavioral adaptation (Crossland, Culham, Kabanarou, & Rubin, 2005) and/or neural plasticity over time (Ferreira et al., 2017; Sanda et al., 2018). 

Acknowledgments

Commercial relationships: none. 

Corresponding author: Laure Pisella. 

Email: [email protected]. 

Address: Centre de Recherche en Neurosciences de Lyon (CRNL), Université Claude Bernard Lyon 1, INSERM U1028, CNRS UMR5292, Bron 69500, France.