A preferred retinal location of fixation can be induced when systematic stimulus relocations are applied

Abstract

Patients with central vision loss obtain visual information by fixating on an object eccentrically with a preferred retinal locus of fixation (PRL). Patients do not always choose the most efficient PRL position, and as a consequence, visual performance is not always fully exploited. This study investigates whether PRLs can be induced by applying systematic stimulus relocations. The PRL was trained using a central scotoma simulation in 15 healthy subjects. They performed different visual tasks during four sessions, after which their reading performance was evaluated. In five subjects the stimulus was relocated to the left hemifield whenever a saccade would place the stimulus on the opposite hemifield. In five different subjects the relocation was inversed: The stimulus was located in the right hemifield. The relocation was 7.5° of visual angle and it was applied horizontally. Five additional subjects naturally chose the PRL location. They were used as the control group to evaluate the development of a PRL. After training, subjects performed visual search tasks on static stimuli. Evaluation after training showed that systematic stimulus relocations can be used to influence the development of the PRL. These results might be significant for the development of training strategies for the visually impaired.

Introduction

When performing daily tasks like reading, walking, or face recognition, healthy humans bring a target of interest onto the fovea with a saccade. Patients with central field loss have to develop strategies to compensate for the lack of foveal input. Since the field of view is restricted to nonfoveal vision, they use a nonfoveal retinal location to refer their saccades and fixations to. This nonfoveal location acts as a pseudofovea and allows patients to acquire the relevant visual information (Cummings, Whittaker, Watson, & Budd, 1985; Fletcher, & Schuchard, 1997; Fuchs, 1992; Guez, Le Gargasson, Rigaudiere, & O'Regan, 1993; Mainster, Timberlake, Webb, & Hughes, 1982; Schuchard, 2005; Von Noorden, & Mackensen, 1962; White & Bedell, 1990). The location is referred to as preferred retinal locus (PRL) and defined to be “one or more circumscribed regions of functioning retina, repeatedly aligned with a visual target for a specified task that may also be used for attentional deployment and as the oculomotor reference” (Crossland, Engel, & Legge, 2011). 

The mechanism underlying the selection of the PRL location is not fully understood. Cheung and Legge (2005) hypothesized three selection categories: function-driven selection, performance-driven selection, and retinotopy-driven selection. The function-driven selection suggests that PRL locations may be determined by the nature of the visual task—for example, a PRL located on the lower visual field is preferable for English reading. On the other hand, the performance-driven selection suggests that the PRL will be either located at the undamaged retinal location with the highest visual acuity or, on the basis of visual attention, the selection will be made in regions with high attentional performance due to the enhancement of visual performance in those regions. Finally, the retinotopy-driven selection suggests that PRL selection might be a consequence of retinotopic reorganizations, where deafferented V1 neurons spontaneously remap to the inputs from retinal locations near the scotoma. Independently of the governing mechanism, the location in which the PRL develops may not always be the most efficient one. In a reading task, studies have demonstrated theoretical and experimental advantages of locating the PRL at the lower region of the visual field among other areas (Chung, Legge, & Cheung, 2004; Deruaz, Whatham, Mermoud, & Safran, 2002; Frennesson & Nilsson, 2007; Guez et al., 1993; Petre, Hazel, Fine, & Rubin, 2000; Whittaker & Lovie-Kitchin, 1993). However, there is a comparable or higher prevalence to locate the PRL on the left field rather than on the lower field in patients with central scotoma (Cummings & Rubin, 1992; Fletcher & Schuchard, 1997; Fletcher, Schuchard, Livingstone, Crane, & Hu, 1994; Sunness, Applegate, Haselwood, & Rubin, 1996). As a consequence, the visual performance can be affected. 

A central vision loss can be simulated in healthy subjects, and the nature of eccentric viewing can be studied (Aguilar & Castet, 2011; Bertera, 1988; Cornelissen, Bruin, & Kooijman, 2005; Fine & Rubin, 1999; Henderson, McClure, Pierce, & Schrock, 1997; Kwon, Nandy, & Tjan, 2013; McIlreavy, Fiser, & Bex, 2012; Scherlen, Bernard, Calabrese, & Castet, 2008; Sommerhalder et al., 2003; Walsh & Liu, 2014). Healthy subjects under central vision loss simulation develop a PRL and suppress normal refoveating saccadic behavior in favor of this location. Furthermore, the development of a PRL is spontaneous and rapid (Kwon et al., 2013; Pidcoe & Wetzel, 2006). Previous studies have also demonstrated that with different training procedures, a new or more favorable PRL can be used by patients with central vision loss (Nilsson, Frennesson, & Nilsson, 2003) or by normally sighted subjects (Lingnau, Schwarzbach, & Vorberg 2008). The present study differs from these studies due to the fact that the inducement is neither confined to a narrow retinal area nor to a single visual task (reading task). In addition, the induced PRL is guided since the early stages of its development and is based on systematic stimulus relocations. 

Methods

Apparatus

Data acquisition was carried out using a gaze-contingent setup based on MATLAB, the Psychtoolbox (Brainard, 1997; Kleiner, Brainard, & Pelli, 2007), the Eyelink toolbox (Cornelissen, Peters, & Palmer, 2002), the Eyelink 1000 Plus eye tracker (SR Research, Ltd., Ontario, Canada), and a ViewPixx/3D display with a vertical refresh rate of 100 Hz and a spatial resolution of 1920 × 1080 pixels. 

Vertical and horizontal positions of the right eye were recorded at 1 kHz while the left eye was patched. To simulate the central scotoma, a gaze-contingent mask was presented at the momentary eye position. The mask was round, with a radius of 3° of visual angle. It was presented in front of a light gray background. The presentation of the scotoma at the momentary eye position was temporally delayed by less than 20 ms after the detection of the eye's position. A chin rest was used to prevent head movements and to locate the eyes at a fixed position 62 cm from the display. 

Participants

The study was performed in accordance with the declaration of Helsinki. Fifteen participants took part in the study: five males and 10 females aged between 24 and 33 years (mean 26.6 years). Thirteen subjects were naive to the purpose of the study and the other two were authors who participated in the control group. Subjects were eye-healthy and had normal or corrected-to-normal visual acuity. 

Study design

The experiment consisted of four training sessions in which subjects had to solve a visual task and subsequently perform a reading task. The main task during the training procedure was to discriminate compound stimuli presented at varying positions on the screen. 

Each subject was randomly assigned to either the left-induced PRL or right-induced PRL or control group. In the induced groups, the training was performed under central vision loss simulation. In addition, a stimulus relocation function was applied to discourage one of the two hemifields. The relocation depended on the momentary gaze and stimulus position. If a saccade located the center of mass of the stimulus on the opposite side of the inducement (between the edge of the scotoma and a distance of 2.5° of visual angle from the edge of the scotoma), the stimulus was relocated on the intended induced hemifield. Consequently, the stimulus was drawn in its new position in the next frame. Figure 1A describes the inducing mechanism for the subjects from the left-induced group. If the subjects intended to locate the stimulus on the right hemifield, the relocation function shifted it to the left hemifield. On the contrary, Figure 1B shows the same procedure for subjects from the right-induced group. The relocation was always applied horizontally into the opposite hemifield and had a constant displacement value of 7.5° of visual angle relative to the stimulus's center of mass. Using this procedure most relocations would let the stimulus reappear in the opposite hemifield, but in some situations the stimulus would actually disappear because it would be shifted to a location within the scotoma. Both situations prevent the usage of the discouraged hemifield. In the control group shown in Figure 1C, the training was performed under central vision loss simulation and no changes were applied to the stimulus position. Thus while performing the task, control subjects were able to locate the stimulus at any desired position outside the scotoma for eccentric fixation. This group was used as a reference to compare the development of new oculomotor strategies to the induced groups and analyze potential effects of the inducing procedure on the development of the PRL. 

Figure 1

 

(A) The left-induced group. Subjects performing a saccade might locate the target in the discouraged semicircular area circumscribed by the dotted line (dotted only for demonstration). In this situation, the stimulus is shifted to the left half of the visual field. Within all other regions, subjects can freely locate the stimulus in the left half of the visual field. (B) The right-induced group. In this situation, the stimulus is shifted to the right half of the visual field when a saccade locates the stimulus in the discouraged semicircular area circumscribed by the dotted line. (C) In the control group, subjects perform a saccade to choose the location in which the stimulus is located.

Stimuli

Every phase of the experiment was conducted in complete darkness. The stimulus consisted of a foveally presented scotoma and a static discrimination target. The simulated scotoma was in a circular shape with ± 3° of visual angle and was colored in dark gray. The discrimination targets were designed to cause a long fixation time, thus increasing the oculomotor learning. 

The background of the screen was light gray in color with a luminance of 64 cd/m². To avoid fixations outside the screen, the location of the stimulus changed randomly within a window of 42° × 21° of visual angle centered on the screen (the screen size being 48° × 27° of visual angle). The overall size of the composed stimulus was 1.7° × 1.7° of visual angle. 

Given that the discrimination targets were big enough to be identified at distances of 3° of visual angle relative to the fovea, the procedure might have become monotonous and unchallenging after two long sub sessions. Therefore, and since crowding decreases the performance during eccentric viewing of a stimulus (Wallace, Chiu, Nandy, & Tjan, 2013), the complexity of the task was increased by adding more components to the composed stimulus, which kept the subjects alert and challenged. 

Figure 2 shows an example of the composed stimuli presented in each session and sub session during the performance of the visual task. In Sessions I, II, and III, the position of the stimulus changed at every trial. 

Figure 2

 

Examples of stimuli presented in each sub session. In Session I colored dots were presented in a random spatial arrangement and subjects had to judge whether there were more red than blue dots. In Session II, a set of vertical lines and squares were presented and subjects had to distinguish between the different shapes. In Session III, horizontal and vertical lines were presented and subjects had to distinguish between the different orientations. In the multiple stimuli session, numbers and letters were presented. For simplification, only one example of each stimulus is presented; however, an example of the complete screen can be found in Figure 8.

In Session I, colored dots were randomly distributed around a predetermined center of mass. In each sub session a new dot and color were added. Subjects had to differentiate between red and blue dots by reporting whether there were more red or blue dots using the up and down arrow keys accordingly. 

In Session II, a stimulus composed of squares and lines was presented. In every sub session, a new component was added. The components were randomly assigned to be either squares or lines. The task was different in every sub session. For example, in Sub session 1, subjects had to report whether the components of the stimulus were the same or different, and in Sub session 4, subjects had to report whether there were more or fewer squares than lines. In addition, during Session II and subsequent sessions, subjects had to press the space key, causing an internal function to randomly select one or more components of the stimulus and mark this selection red (this action is repeated until required component/s were marked red). For example, in Sub session 1, if the components of the stimulus were identical, then subjects had to mark both components red and if not, subjects had to mark only the square red. From Sub sessions 2 to 4, subjects had to mark all squares red. 

In Session III, vertical lines were presented instead of squares and the task was the same as in Session II. 

In the multiple stimuli training, a set of targets were presented simultaneously (numbers or letters). In Sub session 1, two random digits (from 1 to 9) were presented inside a ring. For simplification, only one stimulus is shown in Figure 2. The horizontal positions of the stimuli were −12° and +12° of visual angle relative to the center of the screen, while their vertical position varied randomly between −6.7°, 0°, and +6.7° of visual angle relative to the center of the screen. The stimulus size was approximately 1° of visual angle. Again, the subjects had to press the space key, causing an internal function to randomly select one or more components of the stimulus and marked this selection red (this action is repeated until required component/s were marked red). The task was to mark the digit with the highest value. In Sub session 2, two simple arithmetic operations (addition or subtraction) were presented. The position of the two operations and the size of each digit were the same as in Sub session 1. Again for simplification, only one arithmetic operation is shown in Figure 2. Subjects had to solve the operation and use the space key to mark the operation with the highest solution as red. In Sub session 3, the arithmetic operation was presented in the center of the screen. The size of the digits was 1.5° of visual angle. Additionally, four numbers representing a solution were shown. Their positions (pos_n [x, y]) relative to the center of the screen were pos₁ = (12, 6.7) degrees of visual angle, pos₂ = (12, −6.7) degrees of visual angle, pos₃ = (−12, 6.7) degrees of visual angle, and pos₄ = (−12, −6.7) degrees of visual angle. Their sizes were 1° of visual angle. Figure 2 shows an example of the arithmetic operation at the center of the screen and one of the possible solutions. The task was to calculate the solution of the arithmetic operation and find it among the four numbers. Subjects had to press the space key until the correct answer was marked red. Finally, in Sub session 4, a group of three letters were shown at the center of the display. The letters were 1.5 degrees of visual angle in height. Additionally, four letters were presented at the same position as the numbers in Sub session 3. Three of the four letters were identical to the ones in the center and one was different. Figure 2 shows an example of three letters shown at the center of the screen and one of the four letters shown at the corners of the screen. The task was to press the space key until the different letter was marked red. Figure 7 shows an example of the whole set of stimuli presented in each session and sub session. 

Figure 3

 

Events occurring during a sub session. Subjects had to perform a visual task under central scotoma simulation where a drift correction occurred after every trial. The experiment proceeded in a loop for 10 min. Subsequently the reading performance started, where the subjects had to read with central scotoma simulation to find the correct string of words and answer. Drift corrections were performed after every trial and the reading performance continued in a loop for 2 min.

Figure 4

 

The SDM shows an example of the distribution of the stimulus location relative to the center of simulated scotoma after the performance of a training sub session (Subject 2, Session II, Sub session 3).

Figure 5

 

Example of PRL index and its influence on the PRL value.

Figure 6

 

PRL positions of subjects after the final performance measurement. The two-sample t test applied to subjects from the left-induced PRL versus right-induced PRL results in significant differences, t(8) = −2.88, p = 0.02.

Figure 7

 

Position of the PRL after each training session. Each number represents a subject and the groups are separated by colors and shapes. Squares represent the subjects under the inducing procedure: blue squares for left, red squares for right, and black diamonds for the control group. The gray central region of ±3° of visual angle corresponds to the area covered by the scotoma.

Procedure

Training and reading performance assessment

A 13-point calibration was used at the beginning of the experiment to collect fixation samples from 13 known target points in order to map raw eye data to gaze position at known target positions. Subsequently, a validation with 13 points was performed, which provided information about calibration accuracy. 

Figure 3 shows the events occurring during the experiment. The visual task block includes the central vision loss simulation and the presentation of the compound stimulus. Thereafter, the subject gave a response, and a drift correction was performed, ensuring that the accuracy of the calibration parameters was maintained, and a new trial began. A count-down timer with a starting time of 10 min was turned on during the performance of the visual task. After the time was completed, the experiment then continued with the reading performance. During the reading performance, subjects had to read a string of three words under central scotoma simulation without any inducement. The string of words covered 1.5° × 16° of visual angle and was composed of similar letters in order to enhance the demand of the task, (e.g., WANT WENT WELL). Subjects were asked to read the three words with the central scotoma and press the space key to report successful reading. In this part of the experiment, the stimulus was not relocated, thus also the subjects from the induced groups were able to locate their PRL freely. Subsequently, without central scotoma, subjects were asked to find the string of three words shown previously among two alternatives and press the up or down key to report the answer (Figure 3, answer block). After the subject gave an answer, a drift correction was performed and a new trial started. The measurement of reading performance continued for 2 min. During answer and drift corrections, the timer was paused. 

Final performance assessment

The final performance assessment (FPA) is a measurement of the developed PRL position after the training and without the inducement procedure. The assessment was taken in a separate appointment at the end of the experiment (at least 1 day after the last training session). The analysis was performed to evaluate if the PRL was induced as intended. The visual tasks were identical to the training tasks, except that the function that changed the location of the stimulus was turned off. Thus, subjects from every group were free to choose the PRL during FPA. 

During FPA, subjects performed the fourth sub session of each session for 1.5 min. First, colored dots were shown where the subject had to identify the red dots among blue dots in a five-color stimulus. Second, lines and squares were shown and subject had to report whether there were greater or fewer squares than lines in a five-component stimulus. Third, horizontal and vertical lines were shown and subjects had to report whether there were greater or fewer horizontal than vertical lines in a five component stimulus. Fourth, three letters at the center of the screen were shown and four in each corner of the screen. Subjects had to find, at one of the four screen corners, the letter that was not shown at the center. 

Data analysis

Fixational behavior was evaluated from all gaze data collected during visual task performance. The beginning and end of fixations and blinks were obtained by applying the internal eye tracker criteria. According to these, fixations corresponded to events in which the saccade velocity was below the threshold of 30°/s and blinks corresponded to periods of data where the pupil was undetected. Blinks and saccades were then eliminated from the data. 

To quantify the position and the development of the PRL, the data (horizontal and vertical position components of the eye on the display) was translated to the origin of a Cartesian coordinate system located on the two-dimensional image space. The stimulus position, saved after each trial, was recalculated relative to this origin and was also translated to this Cartesian system. The result obtained depicted the distribution of the stimulus position relative to the gaze (or center of simulated scotoma). It will be referred to as stimulus distribution map (SDM; Figure 4). 

Position of the PRL

The position of the PRL is defined as the spatial location in which the highest density of the SDM is found (Kwon et al., 2013). The density was obtained using a bivariate Gaussian kernel estimator (Botev, Grotowski, & Kroese, 2010). In Figure 4, the small black cross on the red region of the SDM shows the position of the PRL for that case. 

PRL value

A PRL value was introduced to track and quantify the PRL development in detail using the SDMs. Thus, PRL value allows an assessment whether the presented paradigm affected the development of the new oculomotor strategies and a detailed comparison of PRL development in all three groups. 

This value is a combined indicator of the three most important PRL features: PRL concentration, sphericity, and location. Thus, it depended on three criteria: the concentration of the distribution (CI), the index of symmetry of the distribution (SI), and the ratio quantifying the fraction of fixations placing the target out of the scotoma relative to the total fixations (R). To each of the three criteria, a value between 0 and 1 was assigned and the PRL value was calculated for each sub session using Equation 1. Within the PRL value, the fraction of fixations out of the scotoma R integrated the effectiveness of an eccentric fixation together with the shape of the PRL, (SI + CI) / 2. The PRL value ranged between 0 and 1, where values close to 1 represent a very narrow and rounded distributions, located out of the scotoma.  

 

 

To calculate SI, a principal component analysis was used to obtain the coefficients of the longest and shortest components of the distribution map and to calculate their lengths. The length of the components was calculated by taking the 15th and 85th percentile, and the SI was obtained by dividing the shortest dimension over the longest dimension (Cherici, Kuang, Poletti, & Rucci, 2012). With this approach, SI values ranged from 0 to 1, where an SI of 1 represents a circular distribution. 

The CI was used to quantify the concentration of the data around its positional mean (horizontal and vertical). In other words, it is a measure of the concentration of data around its center of mass. To calculate it, the 60th percentile of the distance between every point of the SDM and their mean in x and y was calculated (C). This quantity, in visual angle, was normalized and transformed to index values (between 0 and 1) by means of a linear equation (Equation 2). The normalization factor β was selected to be the highest value that C took among all subjects and sessions could take. With this approach, one assigns values close to 1 for high concentrated SDMs and values close to 0 for low concentrated stimulus distributions maps.  

 

 

The ratio (R) quantifies how many fixations placed the target outside the scotoma. It was calculated by dividing the number of times that the stimulus was located out of the scotoma n_out over the total number of times that the stimulus was fixated n_total. Additionally, the lowest limit of the ratio factor (0) was assigned to be the point in which 50% of the fixations are out of the scotoma and 50% inside the scotoma (Equation 3). With this approach, values close to 1 represented efficient oculomotor behavior, which localized the stimulus out of the scotoma, values close to 0 represented oculomotor behavior, which by chance located the stimulus inside or outside of the scotoma, and values below 0 represented fixations that located stimulus mainly in the scotoma region. In that case, R was assigned to be 0 instead of the negative value. As a consequence, the PRL value was also 0.  

 

 

Figure 5 shows three examples of different fixational behavior with their respective indexes and PRL values. On the left is a subject who tried to foveate the stimulus. The symmetry and concentration indexes are relatively high, but the low R ratio leads to a low PRL value. The figure in the center shows an example case of a subject who started to fixate eccentrically. In this case the CI and SI decreased due to the elongation of the fixational pattern; however, the ratio that quantifies the eccentric fixations increased. This leads to a higher PRL value. Finally, on the right, a subject with a trained PRL is shown. In this case, the stimulus is repeatedly fixated out of the scotoma on a location that was consistently selected, leading to a high PRL value. 

Results

Position of the PRL

PRL position after FPA

In the FPA the stimulus relocation function was turned off and subjects performed four visual tasks with static stimuli. The SDM obtained after the performance of the four visual task was obtained for each subject. To calculate the position of the PRL, the point of peak density of the SDMs was calculated and the results for every subject are presented in Figure 6. 

The horizontal component of the PRL from the left and the right-induced group differ significantly in a two-sample t test, t(8) = −2.88, p = 0.02. The mean horizontal PRL position of subjects from the left-induced group was −2.3° ± 1.2° of visual angle and for the subjects from the right-induced group was 0.3° ± 0.8° of visual angle. Subjects from the control group showed a mean horizontal PRL position of −1.4° ± 2.2° of visual angle. 

Furthermore, all subjects but two showed a PRL located outside of the scotoma, but still in proximity to the scotoma, with a distance between the PRL location and the edge of the scotoma below 3° of visual angle. The two subjects that presented the PRL inside the scotoma (Subjects 12 and 13) alternate a PRL position between two locations (outside and inside the scotoma, Supplementary Figure S1). 

Monitoring the PRL position after each training session

The PRL position was evaluated after 40 min of recording, equivalent to the time taken to record four sub sessions. This enabled a detailed analysis of the PRL development throughout the training. To obtain the position of the PRL, the point of peak density of the SDMs were calculated and the results for every single subject are presented in Figure 7. 

In Session I, nine subjects located the PRL in the center of the scotoma showing that, at the beginning of the training, subjects tried to gaze with the fovea repeatedly. But by Session II, subjects already fixated eccentrically. 

Notice that some PRLs seem to be located inside the scotoma: This is actually an artefact of two PRLs or distributed SDMs in this specific session. Nonetheless, clear PRLs outside of the scotoma are available for every subject in a late phase of training and can be found in Supplementary Figure S2. 

PRL value

Figure 8 shows the SDMs of a sample subject for the complete training procedure with the corresponding performed tasks and PRL index values. Notice that during Session I the subject brought the stimulus from the region of the scotoma to the region outside, and during Session II the stimulus was located mainly out of the scotoma. Note that, in Session II the subject located the stimulus at two positions, but in Session III, only one PRL remained. In the last training session, where multiple stimuli were presented, the SDMs are broader. 

Figure 8

 

SDMs for sample subject 3 (control group). Each SDM plotted with its corresponding sub session is the result of 10 min of recording (while the timer was running). In addition, the indices used to calculate the PRL values are shown with their respective PRL value for every session. Note that the PRL value increased from session to session during the presentation of a single stimulus.

To analyze whether the training improves the oculomotor behavior and whether the paradigm affects the development of new oculomotor strategies, the PRL value was analyzed. 

Figure 9 shows the mean PRL values for every subject of each group as a function of the sub session number. The blue shaded area corresponds to the mean PRL and standard deviations of the left-induced group, the red shaded area to the right-induced group, and the gray shaded area to the control group. “Single stimulus” corresponds to the PRL values collected during the performance of the first 12 sub sessions (or first three sessions). “Multiple stimuli” corresponds to the four sub sessions performed in the multiple stimuli session. During the performance of the single stimulus task, the PRL value appears to increase with training in every group. To test whether the improvement is significant, a paired t test was performed between the first sub session (Sub session 1) and last sub session (Sub session 12) for every group independently. PRL values increased significantly in the right-induced group, t(4) = −2.55, p = 0.004, and control group, t(4) = −14.39, p = 0.0007, showing a successful training. In the left-induced group the PRL value did not increase significantly, t(4) = −2.55, p = 0.062. This might be due to the fact that the variance in final PRL values was high in this group. During the multiple stimuli training, the PRL values dropped. 

Figure 9

 

PRL values as a function of the training sub sessions. The PRL value increases along the performance of the single stimulus training and drops during the performance of the multiple stimuli session.

Additionally, to see whether the paradigm affects the development of new oculomotor strategies, two-sample t tests were performed between the groups. The results showed significant differences between the induced groups, t(22) = 2.64, p = 0.01, and between right-induced and control group, t(22) = 2.65, p = 0.01, but no significant differences between left-induced and control group, t(22) = −0.80, p = 0.42. These results indicate that inducing the PRL on the right hemifield might require longer training time. On the other hand, significant differences were found between the beginning and end of the training in this group, indicating that the training is improving the general performance. 

Regarding the multiple stimuli session, no statistically significant differences were found between the groups (induced groups t[6] = −0.19, p = 8.85; right-induced versus control group t[6] = −0.96, p = 0.37; left-induced versus control t[6] = 0.58, p = 0.58. 

Reading performance assessment during training

During the experiment, the reading performance under central scotoma simulation was assessed with a reading task at the end of each training sub session. A group of three words composed of four letters was presented and the time spent to read the group of words was evaluated. In Figure 10 the mean elapsed time per trial with its respective standard error is shown as a function of the training session for the three groups. 

Figure 10

 

Mean elapsed time per trial as a function of the session number obtained during reading performance measurement. Each session number is divided into four values, corresponding to each sub session. Values at the beginning and end of training were significantly different for the left-induced group, t(4) = −5.69, p = 0.004, right-induced group, t(4) = −3.93, p = 0.01, and control group, t(4) = −8.40, p = 0.001.

The mean elapsed time at the beginning and the end of the training were tested separately in every group and on the reciprocal for equal variances. Results show significant improvements between beginning and end reading time for the left-induced group, t(4) = −5.69, p = 0.004, the right-induced group, t(4) = −3.93, p = 0.01, and the control group, t(4) = −8.40, p = 0.001. 

Initially, subjects in the left-induced group read a mean of 34.4 ± 9.8 words per minutes (wpm), subjects of the right-induced group a mean of 15.7 ± 7.7 wpm, and subjects in the control group read a mean of 30.0 ± 12.5 wpm. After training, subjects of the left-induced group increased their reading speed to 101.5 ± 24.3 wpm, of the right-induced group to 74.0 ± 11.9 wpm, and subjects of the control group to 106.4 ± 15.7 wpm. 

Discussion

The PRL position

To answer the question of whether the location of the PRL can be induced at early stages of its development using systematic stimulus relocation, the location of the developed PRL was evaluated. After the inducement, once the stimulus was no longer relocated, all subjects from the left-induced group placed the stimulus consistently on the left half of their visual field, and three subjects from the right-induced group placed the stimulus on the right half of their visual field. Two subjects from the right-induced group had difficulties to develop a PRL on the right hemifield. Maybe the higher prevalence to locate the PRL on the left hemifield played a role on this difficulty (Cummings & Rubin, 1992; Fletcher et al., 1994; Sunness et al., 1996). 

Previous studies have demonstrated that the PRL can be trained on normally sighted subjects (Lingnau et al., 2008). The presented study demonstrates that the PRL can also be induced to be at a specific hemifield based on systematic stimulus relocation. 

Additionally, our findings show that eight subjects located the PRL below the scotoma, four subjects above, and three subjects to the left of scotoma. The fact that the majority located the PRL below the scotoma agrees with studies showing a higher tendency to locate the PRL on the lower side of the scotoma (Fletcher & Schuchard, 1997). As a PRL located on the lower visual field is better for English reading (Nilsson et al., 1998, 2003), these results support the function-driven selection hypothesis for the development of a PRL, which predicts that PRL positions depend on the visual task. Moreover, most of the subject-developed PRL positions close to the edge of the scotoma (distance less than 3° of visual angle). These results agree with the study from Fletcher and Schuchard (1997), which showed that in 883 eyes with different forms of maculopathy, 88.7% of the PRLs were within 2.5° from the border of the scotoma. Additionally, Sunness et al. (1996) found that among 27 eyes with dry age-related macular degeneration and eccentric PRLs, the PRLs were always within 2° from the scotoma border. These findings support the retinotopy-driven selection mechanism for the development of a PRL, which predicts the PRL at the border of the central scotoma (Cheung & Legge, 2005). 

The systematic stimulus relocation presented in this study can be tailored to the intended PRL location and thus be used to encourage other regions of the visual field. For example, confined regions of the visual field can be selected to induce PRLs. Potential encouraged regions could be narrower, such as regions at the left and right visual field quarters, or circular regions at any part of the visual field. 

PRL development

In correspondence with the previous findings, healthy subjects learned to fixate a target eccentrically within 2 training hours and their behavior under simulated central vision loss showed a spontaneous and fast plasticity that can be attributed to oculomotor learning (Kwon et al., 2013; Pidcoe & Wetzel, 2006). This is in contrast to the clinical observations that imply lengthy adjustment periods in patients with central vision loss (Crossland, Culham, Kabanarou, & Rubin, 2005; White & Bedell, 1990). However, a previous study demonstrated that older adults were slower and used excessive eye movement during a search task and during a central vision loss simulation (Kwon et al., 2012). Thus, future training procedures might have to be adjusted for the patient's age. 

The PRL values during the first three training sessions increased gradually, suggesting that training improves the oculomotor behavior under scotoma simulation when single stimuli are presented. However, differences on PRL values between the right-induced group and the other two groups were observed: Two subjects from the right-induced group showed central fixations and a slower development of PRL at the FPA. These differences might be explained by the large incidence to locate the PRL on the left side of the scotoma in patients with central vision loss (Cummings & Rubin, 1992; Fletcher & Schuchard, 1997; Fletcher et al., 1994; Sunness et al., 1996) and suggest that the inducement in regions with low incidence might require an extra effort in the development of a PRL. Moreover, Liu and Kwon (2016) used a gaze-contingent simulated scotoma to induce a reliable PRL on the left, right, above, and below the scotoma. However, the training time used in their study was between 6 to 7 hr. In our study, subjects were a maximum of 2.6 hr under the training procedure. These results suggest that the duration of the training might play an important role on the development of reliable PRLs. 

The difference in PRL values obtained during the multiple stimuli session might be paradigm induced. The subjects had to perform two different perceptual tasks, which might have required different oculomotor behavior (the first task was to find the correct answer and the second task was to mark the correct answer red). First, to find the correct answer, subjects needed to approach the stimulus and look at each of them eccentrically. Second, since parafoveal color vision does not differ in essential characteristics from foveal color vision under high retinal illumination (Gilbert, 1950), subjects were able to place the scotoma in a central position on the display and still see changes in the color on the stimuli when the space key was pressed. This might have altered the fixational behavior significantly. Consequently, the data used to calculate each index of the PRL value, which was always obtained by the transformation of stimulus position relative to center of scotoma's position, might have been reduced because of the time that the subject spent locating the scotoma at the central position of the screen. Therefore, further studies should consider the use of multiple and colored stimuli in their paradigm. Alternatively, the change from a single stimulus paradigm to a novel multiple stimuli paradigm might have impeded the transfer of the oculomotor behavior, which might have led to the decay of the PRL values. 

Reading performance during training

Subjects improved their reading speed in a similar way in all groups. The left-induced group showed a mean improvement of 67.1 wpm, the right-induced group of 58.3 wpm, and the control group of 76.4 wpm. A comparable improvement was demonstrated by patients with PRL location initially located on the left field of view, which was then moved above or below the central scotoma. Reading speed for those patients showed an improvement from 9 ± 5.8 wpm to 68.3 ± 19.4 wpm (Nilsson et al., 1998). 

Conclusions

This study demonstrates that the location of the PRL can be induced at an early stage of its development using systematic stimulus relocation and that this new paradigm does not impair the PRL development. In addition, the procedure confirmed that normally sighted people can develop the PRL in a fast and spontaneous way. This serves as a starting point for guiding the PRL formation in individuals suffering from visual impairments. 

Acknowledgments

Funding received from Eberhard-Karls-University Tuebingen (ZUK 63) as part of the German Excellence initiative from the Federal Ministry of Education and Research (BMBF). This work was done in an industry-on-campus-cooperation between the University Tuebingen and Carl Zeiss Vision International GmbH. M. Barraza-Bernal is scientist at the Unviersity Tuebingen, K. Rifai and S. Wahl are employed by Carl Zeiss Vision International GmbH and are scientists at the University Tuebingen. 

Commercial relationships: none. 

Corresponding author: Maria Barraza-Bernal. 

Email: . 

Address: Institute for Ophthalmic Research, University of Tuebingen, Tuebingen, Germany.