Monocular eye-cueing shifts eye balance in amblyopia

Abstract

Here, we investigate the shift in eye balance in response to monocular cueing in adults with amblyopia. In normally sighted adults, biasing attention toward one eye, by presenting a monocular visual stimulus to it, can shift eye balance toward the stimulated eye, as measured by binocular rivalry. We investigated whether we can modulate eye balance by directing monocular stimulation/attention in adults with clinical binocular deficits associated with amblyopia and larger eye imbalances. In a dual-task paradigm, eight participants continuously reported ongoing rivalry percepts and simultaneously performed a task related to the cueing stimulus. Time series of eye balance dynamics, aligned to cue onset, are averaged across trials and participants. In different time series, we tested the effect of monocular cueing on the amblyopic and fellow eyes (compared to a binocular control condition) and the effect of an active versus passive task. Overall, we found a significant shift in eye balance toward the monocularly cued eye, when both the fellow eye or the amblyopic eye were cued, F(2, 14) = 27.649, p < 0.01, ω² = 0.590. This was independent of whether, during the binocular rivalry, the cue stimulus was presented to the perceiving eye or the non-perceiving eye. Performing an active task tended to produce a larger eye balance change, but this effect did not reach significance. Our results suggest that the eye imbalance in adults with binocular deficits, such as amblyopia, can be transiently reduced by monocularly directed stimulation, at least through activation of bottom-up attentional processes.

Introduction

Recently, we developed a paradigm using continuous reports of binocular rivalry to measure ongoing changes in eye balance in adults with normal vision (Wong, Baldwin, Hess, & Mullen, 2021). Using this approach, we found that directing a cue stimulus to one eye shifts eye balance toward the cued eye. The cue altered eye balance regardless of whether attention was directed to the perceiving eye or the non-perceiving eye during rivalry. Moreover, because this shift in eye balance was enhanced by a higher attentional load, it appears to be influenced by bottom–up stimulus-driven attention (Zhaoping, 2008) with a contribution from top–down cognitive-driven attention (Zhang, Jiang, & He, 2011). Given that monocular attention can shift eye balance in people with normal and relatively balanced eyes, here we asked whether eye balance in amblyopic adults, who have large, longstanding binocular imbalances due to on-going suppression, can be manipulated in a similar manner. 

A better understanding of how stimulus/attentional modulation can affect the suppression of the amblyopic eye not only has theoretical interest but also has potential therapeutic applications. There is current support for the notion that the primary deficit in amblyopia is the binocular deficit, and the loss of acuity is a secondary consequence (for a review, see Hess, Thompson, & Baker, 2014). Suppression lies at the heart of the binocular deficit. Its reduction is key to the restoration of full binocular function (Hess, 2022), and attention has been implicated as an important factor for understanding suppression (Hou, Kim, Lai, & Verghese, 2016; Hou & Nicholas, 2022; Verghese, McKee, & Levi, 2019). 

In this study, we addressed three related questions. First, in amblyopia, can monocular stimulation (cueing) produce shifts in eye balance, similar to those previously found for normal vision using our binocular rivalry paradigm (Wong et al., 2021), potentially demonstrating a therapeutic role for monocular stimulation/attention in amblyopia? The dichoptic gaming approach developed by To et al., (2011) is based on separate dichoptic stimulation of the fellow and amblyopic eyes within a task in which both sources of information are necessary to play the game. Others (e.g., Hou & Nicholas, 2022) have used a similar approach but with an emphasis on monocular cueing to specifically target top–down attention. Ooi, Su, Natal, and He (2013) showed that cueing can be effective in modulating eye balance in onset rivalry. Therefore, we proposed (hypothesis 1) that monocular cueing can also shift the interocular balance in amblyopia in a steady-state rivalry task. 

Second, we asked whether the modulation of eye balance by monocular stimulus cues is similar between fellow and amblyopic eyes and does it depend on whether the cue is presented to the eye perceiving the rivalrous percept or to the suppressed, non-perceiving eye? In other words, is the cueing susceptibility of the amblyopic and fellow eyes similar, independent of which eye is being suppressed? According to the attentional hypothesis of suppression (Hou & Nicholas, 2022), the effects of cueing might be expected to have a differential effect for the amblyopic eye when it is being suppressed by the sighted eye. This follows from their suggestion that suppression of the response of the amblyopic eye is solely due to a monocular top–down attentional neglect. We proposed (hypothesis 2) that cueing affects the amblyopic eye differently if it is the non-perceiving (suppressed) eye. 

Finally, we asked whether eye balance changes are biased by active versus passive processes—in this case, whether the effect of cue presentation is enhanced by an associated visual task. This bears upon the role of attention in monocular cueing, potentially providing information on the role of top–down feedback, or task-related attention, compared with bottom–up effects that are related to stimulation per se, in the eye-balance shift. We proposed (hypothesis 3) that the monocular attentional effects are not primarily driven by cognitively directed attention (top–down), as there is a major role of stimulus-driven (bottom–up) attentional processes. 

We applied a continuous measure of rivalry dynamics through the use of a joystick. Our approach of describing the time courses of continuous rivalry over extended durations provides a novel opportunity to follow continuous and dynamic changes in eye balance over extended periods of time, not provided when participants’ percepts are reported by button presses, which only indicate discrete periods of exclusive or mixed percepts rather than continuous processes. 

Methods

Participants

Inclusion criteria included the presence of amblyopia or history of amblyopia prior to treatment. Exclusion criteria were the presence of large-angle strabismus (>10 diopters) or an inability to perceive binocular rivalry for stimuli of equal interocular contrast. This latter exclusion criterion was necessary because we required a measurable level of baseline rivalry from which the effects of different cueing procedures could be quantified. Twelve adult amblyopic participants were screened by whether rivalry could be seen when one 300-second run of the experiment was presented, consisting of the presentation of the rivalrous stimuli without cueing (our rivalry-only control condition). Eight participants (four males, four females; ages 27–68 years) perceived rivalry and were eligible to participate in the study. The clinical details of participants with amblyopia are presented in Table 1. All of the strabismic participants we recruited had small-angle strabismus (relative to the width of the stimulus field), and none reported diplopia during testing. Participants provided written informed consent. The experiments were performed in accordance with the tenets of the Declaration of Helsinki and were approved by the Research Ethics Board of the McGill University Health Centre. 

Table 1.

 

Clinical details of subjects with amblyopia or stereo deficits. The strabismic angle was assessed with a major synoptophore. Suppression was assessed with either the Dichoptic Letter Test (Kwon et al., 2014) or the DiCOT suppression test (Baldwin et al., 2024) (see Methods). DS, dioptic sphere; Eso, esotropia; exo, exotropia.

 

Tests of suppression

The Dichoptic Letter Test was adapted from Kwon et al., (2014). This method consists of reading dichoptically displayed Sloan letters, arranged in a layout similar to the Early Treatment Diabetic Retinopathy Study (ETDRS) acuity chart. The letters are bandpass filtered (2 cycles per degree [cpd]), presented on a gray background (60 cd/m²). Two different sets of five letters of different contrast were presented dichoptically and overlapping to each eye. At each position, the identity and interocular contrast ratio of the letter on each chart differs. Participants were instructed to read aloud the more visible letter at each of the five positions in the chart from left to right. The relative contrast of overlapping letters in each eye was adjusted via a customized adaptive procedure. 

For the Dichoptic Contrast Ordering Test (DiCOT), cyan and red sprites representing fruit drawings (i.e., apple, cherry or grape) with normalized root mean square (RMS) contrast were displayed on an iPad Mini display (5th generation, A2133; resolution, 2048 × 1536 pixels; Apple, Cupertino, CA) with a uniform white background (561 cd/m²) with varying contrasts. Participants were asked to select the fruits by tapping the screen in order from highest to lowest contrast. The contrast on each trial was set by an entropy-minimizing procedure that attempted to determine the best value for the interocular contrast ratio in the fewest trials (Baldwin, Lorenzini, Fan, Hess, & Reynaud, 2024) 

Apparatus

Stimuli were generated using MATLAB (MathWorks, Natick, MA) and Psychophysics Toolbox (Brainard, 1997). All stimuli were displayed on an ASUS Desktop PC with a gamma-corrected 23-inch display (ASUSTek Computer, Taipei, Taiwan). An NVIDIA 3D Vision 2 system (NVIDIA, Santa Clara, CA) was used for stimulus presentation. Stimuli were presented using frame interleaving with synchronized shutter glasses. Each eye was presented stimuli at 60 Hz, for a total refresh rate of 120 Hz. The monitor resolution was 1920 × 1080 pixels with a mean luminance of 64 cd/m². The room was dark during testing, with the test screen as the only light source. 

Visual stimuli

Eye balance throughout the experiment was measured using binocular rivalry. Stimuli are illustrated in Figure 1. The two rivalry stimuli were gratings of orthogonal orientations, one presented to each eye oriented at +45° and –45°, with a diameter of 9.9 degrees of visual angle (full diameter at half height) and spatial frequency of 1.26 cpd at 50% contrast. The edges of the grating stimulus were softened with a raised cosine envelope (width = 0.4°). The stimulus size was large relative to the size of the strabismus, ensuring minimal relative change in spatial position during perceptual alternation for any participant (see Table 1). The presentation of the two grating orientations was counterbalanced between the two eyes. Participants viewed the screen at 100 cm, giving a resolution of 60 pixels per degree of visual angle. A binocular fixation ring, 16° in diameter, was placed around all stimuli. Attention was directed to one eye by presenting an additional visual stimulus to that eye, referred to as the stimulus cue. The cue stimulus consisted of a static ring of 12 colored discs. The discs were red, blue, or green, chosen to be easily discriminable at the peripheral location presented (see Figure 1). All cue stimuli appeared intermittently for one second. The delay between cueing stimuli was at least 5 seconds, with an additional delay of x seconds with probability 0.95(x/10). The variable cue intervals helped prevent subjects from predicting the time at which the cue would appear. 

Figure 1.

 

Stimuli presented to the left and right eyes during the experiment. Static grating stimuli were presented, one to each eye, throughout the experiment. This example shows stimuli presented during the left-eye cueing condition where the chromatic attention stimuli were briefly presented monocularly to the left eye for 1-second durations (12 clearly visible colored circles consisting of four red, four green, and four blue). During active cueing, at each chromatic stimuli presentation, there was a 20% chance it was the (A) target (colored circles with both vertical and horizontal symmetry) and 80% chance it was the (B) non-target (colored circles with one of vertical or horizontal symmetry). (C) During passive cueing, colored circles with no symmetry were presented.

Procedure

The experiment software sampled the behavioral responses of the subjects every 100 ms throughout one trial lasting 300 seconds. At all times and in every condition, participants were instructed to continuously report their grating percepts by moving a joystick only in the left–right axis. Horizontal movement to the two extreme positions (left vs. right) represented exclusive grating percepts. Joystick positions in between the two extremes represented varying degrees of mixed percepts relative to its proximity to either extreme. For example, the joystick placed at the center of the two extremes represents a percept that contains an equal amount of left-oblique and right-oblique content. 

In every trial, the grating stimuli were presented continuously throughout the 300-second duration of the trial. The first and last 60 seconds of each trial consisted of rivalry only. In conditions with the cue stimuli (see below), 1-second cueing stimuli appeared intermittently throughout 180 seconds starting from 60 seconds after the start of the trial and finishing 60 seconds before the end. 

We investigated the role of active attention in cueing with a color symmetry judgment task. In addition to reporting ongoing rivalry percepts, participants performed a secondary task using the cue stimuli. The cue stimuli were 12 chromatic discs (0.7° in diameter), presented in a ring surrounding the grating stimulus, 5.6° from the central fixation point (see Figure 1). While maintaining central viewing, participants were asked to covertly monitor the screen during the rivalry task for a brief presentation of the surrounding cue stimulus. The task was to press the button only for target stimuli and otherwise ignore the non-target “catch” stimuli. The target stimuli and non-target catch stimuli differed in the symmetry of the color arrangement: The target consisted of colored circles arranged with vertical and horizontal color symmetry, but non-target catch stimuli lacked this property. For each cue stimulus presentation, there was a 20% chance that a target stimulus would be presented. Audio feedback occurred during button presses and differed between correct and incorrect responses to the cue stimuli. All participants used one hand to both press the button (located on top of the joystick) and maneuver the position of the joystick. Button pressing did not impede the simultaneous horizontal maneuvering of the joystick. For the control, we also ran a passive cueing task in which the peripheral cue was of the same form as for the active task, except the disc arrangement had no color symmetry (see Figure 1C). The subject was asked not to respond to the peripheral cueing stimulus when it occurred, but in all other respects the trials were the same as the active cueing condition. In the rivalry-only (no stimulus cueing) condition, participants only reported their grating percepts. This condition is used as a control to characterize rivalry dynamics that occur when simply viewing the two different gratings, one in each eye, without additional manipulations. 

Four conditions were tested in which cue stimuli were presented monocularly to the amblyopic eye (AE) or the fellow eye (FE), or they were either presented binocularly or not presented (rivalry-only control condition). The experiment was completed over three visits on different days. One block consisted of five repetitions of the 300-second trial all of one condition. Between two different blocks, participants took a 15-minute break. Participants came in on three different visits (different days) to complete all of the conditions tested, as follows. On the first visit, participants were tested on the binocular active cueing condition followed by a 15-minute break, which was followed by the monocular left-eye active condition. The binocular condition always preceded the monocular condition to avoid transfer of any cue-induced eye balance changes across the two conditions. On the second visit, participants were tested on the rivalry-only condition followed by a 15-minute break, which was followed by the monocular right-eye active condition. The rivalry-only condition always preceded the monocular condition for the same reason as above. On the third visit, participants were tested on the binocular passive condition followed by a 15-minute break. A random number generator was used to determine whether the next condition tested was the monocular left- or right-eye passive condition with equal probability. This was followed by a 15-minute break. The remaining monocular left- or right-eye passive condition was then tested. All five repetitions of each condition were completed before a different condition was tested. 

Bootstrapping

The same steps were followed for each non-parametric bootstrapping completed in this paper. For a given dataset, sampling with replacement was done 1000 times. We found the mean for each sample. The bootstrapped sample mean was the mean of the 1000 means from the bootstrapped samples. The 95% confidence interval (CI) of the sample mean was determined by ordering the means of the 1000 bootstrapped samples and taking the values from the 2.5th to 97.5th percentiles. 

Group analyses for a given metric were conducted by bootstrapping from each subject's mean metric. For example, group-averaged curves in the amblyopic eye-cueing condition were found by bootstrapping from each subject's average amblyopic eye-cueing curves. For each bootstrap iteration we selected N averaged data samples from our N subjects (sampling with replacement). 

Results

Task performance

Data for task performance are plotted in Figure 2. Participant performance on target versus catch stimuli trials was evaluated by calculating an “adjusted” hit rate (HR), taking account of false positives, using the following formula: Adjusted HR (corrected for guessing) = TP/(TP + FN) – FP/(FP + TN), where TP is true positive, FN is false negative, FP is false positive, and TN is true negative. Figure 2A shows task performance by participant. Although the wide range of adjusted hit rates indicates that task difficulty varied for each participant, all participants performed above chance (an adjusted hit rate of 0%). Seven of the eight participants had adjusted hit rates that ranged between 65% and 93%, and one participant had an adjusted hit rate of 19%. Participants could perform the task despite its difficulty. Figure 2B shows task performance split by the eye cued. There was no significant difference within participants between the monocular eye-cueing conditions. 

Figure 2.

 

Task performance. Participant performance on target versus catch stimuli trials was evaluated by calculating an “adjusted” hit rate, taking account of false positives, using the following formula: Adjusted HR = TP/(TP + FN) – FP/(FP + TN). (A) Five participants had adjusted hit rates that ranged from 79% to 93%, two participants had adjusted hit rates that ranged from 65% to 67%, and one participant had an adjusted hit rate of 19%. Although the wide ranges of performance indicate that task difficulty varied for each participant, most participants performed well above chance (an adjusted hit rate of 0%). (B) Hit rates within participants did not vary significantly between monocular eye-cueing conditions.

Determining baseline eye dominance

We use an ocular dominance index (ODI) index to characterize participants’ baseline eye dominance. The index is determined from the reports of grating percepts which are represented by joystick movements. As described in Wong et al. (2021), the ODI is calculated by taking the difference of the joystick curve area representing the left-eye percept and the right-eye percept over the sum of the areas. This simplifies to  

\begin{eqnarray*}{\rm{ODI}} = \frac{{{\rm{\Sigma y\ }}}}{{{\rm{\ }} \Sigma \left| {\rm{y}} \right|}}\end{eqnarray*}

where y is the joystick position between the two extreme values, –1 ≤ y ≤ 1, representing exclusive percepts in the left or right eye, respectively. 

Positive ODI values indicate right-eye dominance, and negative ODI values indicate left-eye dominance. Each participant's baseline sensory eye dominance was calculated from averaging the ODI of the first 60 seconds of all 20 trials (five repeats of four conditions). Baseline ODIs can be seen in Figure 3, with error bars giving bootstrapped 95% CIs. Each participant’s dominant eye as measured from their baseline ODIs was consistent with the eye reported as their fellow eye, except for one participant. This participant reported their right eye as their fellow eye, whereas their baseline ODI measurement indicated that the person had relatively balanced eyes. 

Figure 3.

 

Mean baseline ocular dominance indices over all trials for each participant. Positive values represent right-eye dominance, and negative values represent left-eye dominance. Balanced eye dominance is represented by 0. Error bars are the 95% CIs.

Determining eye balance: Time-series analysis

To determine whether cue presentation produced changes in eye balance and to characterize the dynamics of these changes, we transformed the series of joysticks positions (the percept positions) to calculate an eye-balance time series representing the variation in eye balance over time, with time 0 indicating cue onset. Derived from the eye-balance time series, we used two different metrics to characterize changes in eye balance following cue onset: (a) eye-balance shift and (b) modification of rivalry dynamics. 

Eye-balance shift

For each subject and eye-cueing condition, we took the joystick time series and divided them into epochs based on when the individual cue stimuli were presented. We aligned these epochs (taken from all five repeated trials) to the attention stimulus onset and averaged across percept reports to obtain a mean eye-balance time series for each subject × cueing condition (i.e., for each subject, there are three mean eye-balance time series, one for each of FE cueing, AE cueing, and binocular cueing). 

The extent to which participant's fellow eye dominated the rivalry percept varied across subjects. To account for this difference in baseline eye dominance within group analyses, we first normalized the mean eye-balance time series. We did this for each subject by finding the mean baseline eye balance, averaging across all rivalry-only trials, and subtracting this value from the cueing mean eye-balance time series. This resulted in shifting the eye-balance time series so that y = 0 represented the normalized baseline mean eye balance during rivalry. Any fluctuations in rivalry due to changes in fusion were taken into account by normalization to the rivalry-only condition. 

Group-averaged times series were found by bootstrapping the set of participants’ normalized mean eye-balance times series (as described above) and are represented as a curve in Figures 4 and 5. Eye balance is shown relative to the subject's FE with positive values and the AE with negative values. The value y = 0 represents the mean eye balance during rivalry—the greater the change from y = 0, the greater the shift in balance toward one eye. The bootstrapped 95% CIs of the group averaged time series were also found. The maximum change in eye balance after cue onset was found for each set of time series for each subject, and group statistics were obtained on this dataset. 

Figure 4.

 

(A, B) Time series of group averaged and normalized eye balances aligned to attention stimuli onset (t = 0) in the active cueing condition (A) and passive cueing condition (B). The 95% CIs of the group averaged curves are represented by the colored shaded regions. Non-overlapping shaded regions between conditions for any given time indicate a significant difference. The gray-shaded region represents the presentation duration of the attention stimuli. Group averaged curves are categorized by the eye-cueing condition: fellow eye (red), amblyopic eye (blue), and binocular (green). Curves are normalized to the rivalry-only curve (see Methods) such that the y = 0 line represents eye balance during rivalry only. Eye balance is shown relative to each participant's FE with positive values and the AE with negative values. The greater the magnitude, the greater the shift in balance toward one eye. (C) A two-way repeated-measures ANOVA showed a statistically significant main effect for eye-cueing condition on the size of the eye-balance shifts, F(2, 14) = 27.649, p < 0.01, ω² = 0.590; no statistically significant main effect of task on the size of the eye-balance shifts, F(1, 7) = 1.053, p = 0.339, ω² = 0.002; and no statistically significant interaction between task and eye-cueing condition on the size of the eye-balance shifts, F(2, 14) = 3.259, p = 0.069, ω² = 0.069. Post hoc comparisons with Bonferroni correction on eye-cueing condition showed significant differences between cueing the fellow eye and amblyopic eye (p < 0.01; 95% CI of the difference, 0.391–0.844) and cueing the amblyopic eye and binocularly (p = 0.003; 95% CI of the difference, −0.345 to −0.571). There was no significant difference between cueing the fellow eye and binocularly (p = 0.017; 95% CI of the difference, 0.046–0.499). Simple-effect analyses showed a significant difference in the size of eye-balance shift between eye-cueing conditions at both task levels, active cueing (p < 0.001) and passive cueing (p = 0.001). There was no significant difference in size of eye-balance shift between task levels when the fellow eye was cued (p = 0.135), amblyopic eye was cued (p = 0.073), or when cued binocularly (p = 0.641).

Figure 5.

 

Time series of individual non-normalized eye balances aligned to attention stimuli onset (t = 0) in the active cueing and passive cueing conditions, plotted as in Figure 4A and 4B without the normalization. Curves are categorized by the eye-cueing condition: fellow eye (red), amblyopic eye (blue), binocular (green), and rivalry only (orange). Here, the y = 0 line represents eye balance reflecting equal contributions from each eye. Eye balance is shown relative to each participant's FE with positive values and the AE with negative values. The greater the magnitude, the greater the eye balance toward one eye.

A two-way repeated-measures analysis of variance (ANOVA) showed a statistically significant main effect for eye-cueing condition on the size of the eye-balance shifts, F(2, 14) = 27.649, p < 0.01, ω² = 0.590; no statistically significant main effect of task on the size of the eye-balance shifts, F(1, 7) = 1.053, p = 0.339, ω² = 0.002; and no statistically significant interaction between task and eye-cueing condition on the size of the eye-balance shifts, F(2, 14) = 3.259, p = 0.069, ω² = 0.069. Post hoc comparisons with Bonferroni correction on eye-cueing condition showed significant differences between cueing the fellow eye and amblyopic eye (p < 0.01; 95% CI of the difference, 0.391–0.844) and cueing the amblyopic eye and binocularly (p = 0.003; 95% CI of the difference, –0.345 to –0.571). There was no significant difference between cueing the fellow eye and binocularly (p = 0.017; 95% CI of the difference, 0.046–0.499). Simple-effect analyses showed a significant difference in the size of eye-balance shift between eye-cueing conditions at both task levels, active cueing (p < 0.001) and passive cueing (p = 0.001). There was no significant difference in the size of eye-balance shift between task levels when the fellow eye was cued (p = 0.135), amblyopic eye was cued (p = 0.073), or when cued binocularly (p = 0.641). 

Modification of rivalry dynamics

Similar to the eye-balance shift analysis, for each subject and eye-cueing condition we divided the joystick time series into epochs based on when the individual cue stimuli were presented. In the monocular cueing conditions (FE or AE cued), these epochs were further grouped by the eye perceiving at cue onset (i.e., FE perceiving or AE perceiving). In this analysis, the “perceiving eye” and “non-perceiving eye” are defined by the eye that is perceiving or non-perceiving at the time of cue onset. In the binocular condition, every epoch was labeled as binocularly cued. Within each group (perceiving eye, non-perceiving eye, and binocularly cued), we aligned these epochs to the attention stimulus onset and averaged across percept reports to obtain a mean eye-balance time series for each subject × cued eye (i.e., for each subject, there are three mean eye-balance time series, one for each of the perceiving eye cued, non-perceiving eye cued, and binocularly cued groups). 

The steps above were done for the rivalry-only condition, as well. Because there was no cueing in the rivalry-only condition, a set of attention stimulus onset timings was generated and used to conduct the analysis. Each epoch was labeled as “rivalry only” because neither eye was cued. Epochs were aligned to the generated attention stimulus onset and averaged across percept reports to obtain a mean rivalry-only eye-balance time series for each subject. 

The difference in percept dynamics between rivalry-only and cueing conditions was found for each subject by subtracting the mean rivalry-only eye-balance time series from each mean cued eye (perceiving eye cued, non-perceiving eye cued, or binocularly cued) time series. These curves are referred to as the mean difference curves. 

Group-averaged mean difference curves were found by bootstrapping the set of subjects’ mean difference curves and are shown in Figures 6 and 7. Eye balance is shown relative to the subject's perceiving eye with positive values and the non-perceiving eye with negative values. The value y = 0 represents the mean rivalry dynamics without cueing. Larger magnitudes from y = 0 represent larger eye balance differences from rivalry only. The bootstrapped 95% CIs of the group averaged time series were also found. The maximum difference from rivalry dynamics after cue onset was found for each set of time series for each subject, and group statistics were done on this dataset. 

Figure 6.

 

(A, B) Time series of group averaged and normalized eye balances aligned to attention stimuli onset (t = 0) relative to the eye that is dominating the percept (the perceiving eye) at the time of attention stimuli onset in the active cueing condition (A) and passive cueing condition (B). The 95% CIs of the group averaged curves are represented by the colored shaded regions. Non-overlapping shaded regions between conditions for any given time indicate a significant difference. The gray-shaded region represents the presentation duration of the attention stimuli. Group averaged curves are categorized by the eye cued: perceiving eye (red), non-perceiving eye (blue), and binocular (green). Curves are normalized to the rivalry-only curve relative to the dominating eye at cue onset (see Methods) such that the y = 0 line represents dynamics during rivalry relative to the perceiving eye at cue onset. Eye balance is shown relative to each participant's perceiving eye (PE) with positive values and the non-perceiving eye (NPE) with negative values. The greater the magnitude, the greater the shift in balance toward one eye. The four subpanels below show these results when the perceiving eye was the FE or the AE. (C) A two-way repeated-measures ANOVA showed a statistically significant main effect for eye-cueing condition on the size of the eye-balance shifts, F(2, 14) = 54.830, p < 0.001, ω² = 0.708; no statistically significant main effect of task on the size of the eye-balance shifts, F(1, 7) = 0.032, p = 0.863, ω² = 0; and no statistically significant interaction between task and eye-cueing condition on the size of the eye-balance shifts, F(2, 14) = 0.387, p = 0.686, ω² = 0. Post hoc comparisons with Bonferroni correction on eye-cueing condition showed significant differences between cueing the perceiving eye and non-perceiving eye (p < 0.001; 95% CI of the difference, 0.446–0.759), cueing the perceiving eye and binocularly (p < 0.001; 95% CI of the difference, 0.168–0.481), and cueing the non-perceiving eye and binocularly (p < 0.001; 95% CI of the difference, −0.435 to −0.122). Simple-effect analyses showed a significant difference in the size of eye-balance shift between eye-cueing conditions at both task levels, active cueing (p < 0.001) and passive cueing (p < 0.001). There was no significant difference in size of eye-balance shift between task levels when the perceiving eye was cued (p = 0.744), when the non-perceiving eye was cued (p = 0.629), or when cued binocularly (p = 0.869).

Figure 7.

 

Time series of individual non-normalized eye balances, relative to the eye dominating the percept (the perceiving eye) at the time of attention stimuli onset, aligned to attention stimuli onset (t = 0) in the active cueing and passive cueing conditions. Plotted as in Figures 6A and 6B, with the exception of no normalization. Curves are categorized by the eye-cueing condition: perceiving eye (red), non-perceiving eye (blue), binocular (green), and rivalry only (orange). Here, the y = 0 line represents eye balance reflecting equal contributions from each eye. Eye balance is shown relative to each participant's FE with positive values and the AE with negative values. The greater the magnitude, the greater the eye balance toward one eye.

A two-way repeated-measures ANOVA showed a statistically significant main effect for eye-cueing condition on the size of the eye-balance shifts, F(2, 14) = 54.830, p < 0.001, ω² = 0.708; no statistically significant main effect of task on the size of the eye-balance shifts, F(1, 7) = 0.032, p = 0.863, ω² = 0; and no statistically significant interaction between task and eye-cueing condition on the size of the eye-balance shifts, F(2, 14) = 0.387, p = 0.686, ω² = 0. Post hoc comparisons with Bonferroni correction on eye-cueing condition showed significant differences between cueing the perceiving eye and non-perceiving eye (p < 0.001; 95% CI of the difference, 0.446–0.759), cueing the perceiving eye and binocularly (p < 0.001; 95% CI of the difference, 0.168–0.481), and cueing the non-perceiving eye and binocularly (p < 0.001; 95% CI of the difference, –0.435 to –0.122). 

Simple-effect analyses showed a significant difference in the size of eye-balance shift between eye-cueing conditions at both task levels, active cueing (p < 0.001) and passive cueing (p < 0.001). There was no significant difference in the size of the eye-balance shift between task levels when the perceiving eye was cued (p = 0.744), the non-perceiving eye was cued (p = 0.629), or when cued binocularly (p = 0.869). 

In summary, our results have demonstrated, using binocular rivalry, that monocular cueing produces transient shifts in eye balance in amblyopia in both the amblyopic and fellow eye. Both the amblyopic and fellow eye are shown to be susceptible to monocular cueing regardless of whether that eye was suppressed or not at a given time. We did not find statistically significant task load dependency as was found in our previous study of binocularly normal subjects (Wong et al., 2021). We conclude that a cue in the amblyopic eye can lead to more balanced rivalry, even in the absence of a task. 

Discussion

Using a monocular/binocular stimulus cueing paradigm within a binocular rivalry task, we assessed the extent to which the longstanding suppressive imbalance in adults with amblyopia could be dynamically modulated. In eight amblyopic adults, some with strabismus and some with anisometropia, we were able to show that the presentation of a monocular visual stimulus, with or without an accompanying task, could influence the instantaneous eye balance even if presented to the amblyopic eye even though the degree of imbalance was extreme and had been longstanding. Thus, we accept hypothesis 1. 

These results in amblyopes are not that dissimilar to those that we have previously presented for normal binocular observers using the same paradigm (Wong et al., 2021) and by colleagues using a different paradigm (Zhang et al., 2011). Normal observers, such as amblyopes, exhibit changes in instantaneous eye dominance to briefly presented monocular stimuli (i.e., stimulus cueing) that affect both perceiving and non-perceiving eyes. Stimulus cueing presented to the perceiving eye extends its rivalry dominance, whereas if it is presented to the non-perceiving eye it promotes a switch in dominance. In amblyopes, the modulatory effects on eye balance of the monocular stimulus presentation were significant for the amblyopic eye. This modulatory effect for the amblyopic eye did not depend on whether the amblyopic eye was the perceiving eye or the non-perceiving (suppressed) eye. This latter finding runs contrary to hypothesis 2—namely, that cueing affects the amblyopic eye differently if it is the non-perceiving (suppressed) eye). Furthermore, whereas in normal subjects in our previous study we found that there was a significant effect of task load and eye-cueing condition (dominant vs. non-dominant) on the modulation of eye balance (Wong et al., 2021), this was not significant in amblyopes. There was a trend, however, for the number of amblyopies tested, as significance was not reached. Thus, hypothesis 3 can be accepted—namely, that the monocular attentional effects are not driven by cognitively directed attention (top–down), as there is an important role for stimulus-driven (bottom–up) attentional processes. 

Relevance to attention

What is the mechanism responsible for these stimulus modulatory effects on eye balance? This could be explained in two seemingly different ways. The first is in terms of low-level masking effects. It is known that the contrast (at and above threshold) of a stimulus in one eye can be inhibited by a stimulus in the other eye and that this dichoptic masking effect is quite broadly tuned for spatial frequency, orientation, and visual field locus (Meese & Hess, 2004; Meese & Hess, 2005). This may be enough to influence dominance in a binocular rivalry task (Blake, 1977) in a way consistent with the present results; the eye presented with the cue becomes dominant by interocularly suppressing the fellow eye. Such a masking-based explanation would not be expected to show a dependence on task load, again consistent with the present results. However, it would be expected to show some dependence on whether the monocular stimulus was presented to the fellow or amblyopic eye. Dichoptic masking studies have shown that the amblyopic eye is less able to inhibit the fellow eye than vice versa (Zhou et al., 2018), and this does not depend on the mask and stimulus being in spatial register (Huang, Baker, & Hess, 2012). That is contrary to the present findings, in which the monocular stimulus presented to the amblyopic eye initiated a comparable shift in dominance compared to fellow eye presentation. This finding is inconsistent with a purely masking-based explanation. 

An alternate explanation for these results and the one that we favored in interpreting our previous results for normal binocular subjects (Wong et al., 2021) involves visual attention in its most broad context. We leave open the possibility that attentional processes could be both bottom–up (that is, automatically activated by stimulus cue presentation) as well as top–down and directed by cognition, depending, for example, on whether or not the correct stimulus cue was presented. Attentive drive associated with cue presentation, regardless of task, is a potential explanation for our results. The non-significance of task load suggests, at the very least, that a top–down contribution is not the main explanation. 

Relevance for amblyopia

Eye balance is disturbed in amblyopia because of suppressive mechanisms. Under normal binocular viewing, sensory information conveyed via the amblyopic eye is, to a greater or lesser extent, suppressed. To rectify this problem, suppression must be reduced, and when it is normal binocular combination is restored (Mansouri, Thompson, & Hess, 2008). There are two very different explanations for the suppression that lies at the heart of the visual problem in amblyopia: a low-level explanation based on imbalanced binocular inhibitory networks (Hess et al., 2014) and an explanation based solely on a monocular top–down attentional deficit (Hou et al., 2016). According to the first explanation, the monocular cue would be expected to have a greater modulatory effect when presented to the amblyopic eye because the stimulation it provides helps redress the net imbalance in interocular inhibition that, at baseline, favors the fellow good eye. The lack of a significant task load effect would be consistent with an essentially bottom–up stimulus-based facilitation. In terms of the second explanation, there is no evidence from the current task to suggest that there is deficient top–down attentive drive to the amblyopic eye because there appears to be only a weak task-load dependence, one that is comparable between fellow and amblyopic eyes of amblyopes and between binocularly normal subjects and amblyopes. Whether this bottom–up facilitation is considered attentional or non-attentional is somewhat semantic—what it is not is what is traditionally thought of as top–down attention. 

There are more dominant, bottom–up attentional influences (i.e., stimulus-dependent) (Zhaoping, 2008) which by their nature help to compete with the steady-state suppressive imbalance, somewhat akin to reducing the contrast of fellow eye stimulation which previously has been shown to be effective at re-establishing binocular function (Mansouri et al., 2008). There may also be additional attention-based top–down mechanisms that are task dependent that make a small contribution to redressing the imbalance via top–down facilitation. In our present study, we did not find a statistically significant task-load dependency, as we did in our previous study of normal subjects (Wong et al., 2021). However, the present data for amblyopes, using our particular task, suggest a trend in that direction which, had we been able to recruit more amblyopes, might have become significant. Nevertheless, any top–down influence in amblyopia, at least using this paradigm, is not noticeably different from that seen in binocularly normal observers, and for that reason the present results do not support the suggestion that amblyopic suppression is wholly due to reduced attentional drive to the amblyopic eye from higher brain regions (Hou et al., 2016; Hou & Nicholas, 2022) . We note that the data collection for this study was conducted just prior to the COVID pandemic, and recruiting further participants or recalling participants was not allowed. 

Relevance to binocular therapy in amblyopia

The new impetus to redefine amblyopia as a primary loss of binocular vision with a secondary loss of monocular acuity comes from a better understanding of not only the likely etiology but also the greater functional importance of binocular vision (Hess, 2022). Recent dichoptic stimulation using gaming platforms has been successful in restoring both binocular and monocular function (Birch et al., 2015; Kelly et al., 2016; Kelly et al., 2018; Webber, Wood, & Thompson, 2016). Attentional modulation is an important part of this approach as both bottom–up stimulus-based (dichoptic stimulation) and top–down cognitively based (the gaming success requires information from both sighted and amblyopic eyes) components are necessary for a successful outcome (Gao et al., 2021). Recently Ooi and colleagues (Ooi et al., 2013; Xu, He, & Ooi, 2010; Xu, He, & Ooi 2012) developed a complementary but different approach, the push–pull therapy, where monocular attention directed to the amblyopic eye when the other eye is also presented with a spatial target results in reducing the interocular imbalance. The results of the present study are complementary to those of Ooi and colleagues but extend beyond the specifics of their approach. Although we conclude, as did Ooi and colleagues, that attentional modulation can provide therapeutic benefits, our approach is different in that (a) we used a steady-state not a transient presentation, (b) our attentional stimulus was linked with a task so that we could assess the role of top–down attentional load, and (c) we showed the effect of cue presentation not only for the perceiving eye (their case) but also for the non-perceiving eye. The results of the second point allowed us to ascertain whether the attentional benefit was solely due to bottom–up stimulus-driven attention or whether cognitively directed top–down attention also contributed. The additional benefit obtained as a result of the third point involves determining whether, as has been suggested (Hou et al., 2016; Hou & Nicholas, 2022; Ooi et al., 2013; Verghese et al., 2019), amblyopic suppression is solely due to attentional neglect from higher cortical levels. 

In summary, our findings are more consistent with the theory that suppression in amblyopia is based on imbalanced binocular inhibitory networks, in which interocular inhibition can be alleviated by bottom–up monocular stimulus presentation. The contribution of top–down attention appears too small to be appreciated in the present study. Although it remains to be seen whether these binocular inhibitory networks can be permanently altered after prolonged or repeated periods of monocular attention, our findings that eye balance can be shifted transiently toward the amblyopic eye prompts further investigation into the potential therapeutic role of monocular stimulation in individuals with amblyopia. 

Acknowledgments

Commercial relationships: none. 

Corresponding author: Kathy T. Mullen. 

Email: [email protected]. 

Address: McGill Vision Research, Department of Ophthalmology & Visual Sciences, McGill University, Montreal, QC H4A 0A4, Canada.