Attentional Eye Selection Affects Perception During Binocular Rivalry

Abstract

Purpose

When dissimilar images are presented to the two eyes, perception alternates between them, a phenomenon known as binocular rivalry. This study aimed to investigate how attentional eye selection modulates sensory eye dominance across a range of stimulus sizes during binocular rivalry.

Methods

Twelve normally sighted observers (mean age: 34 ± 14 years; six females) viewed dichoptically presented orthogonal gratings with distinct contrast-reversal rates (6 Hz and 7.5 Hz). Binocular rivalry was assessed across five stimulus sizes (1°, 2.5°, 5°, 10°, 15°) under two conditions: without attention and with attention directed to the nonsighting eye, as determined by the hole-in-the-card test. The occurrence of perceptual alterations, the duration of perceptual dominance, and the relative proportion of monocular and mixed percepts were measured.

Results

In the absence of directed attention, the sighting eye exhibited greater perceptual dominance, evidenced by higher occurrence rates, longer dominance durations, and larger dominance proportions. When attention was selectively directed to stimuli of the nonsighting eye, perceptual dominance shifted toward the attended nonsighting eye, with the most pronounced effect observed at the 5° stimulus size. The proportion of mixed percepts remained unaffected by such selective attention across all stimulus sizes.

Conclusions

Voluntary attentional eye selection can modulate sensory eye dominance during binocular rivalry, with the magnitude of this modulation varying with stimulus size.

Translational Relevance

These findings highlight the potential of attentional strategies in modulating sensory eye dominance, offering insight that may inform therapeutic interventions for visual disorders such as amblyopia.

Introduction

Binocular rivalry occurs when dissimilar images are presented to the two eyes, resulting in perceptual competition between the two images. Studies have shown that this perceptual competition during binocular rivalry can be modulated by lower-level stimulus features such as stimulus contrasts^1–3 or spatial frequency^4–7 and also by higher-level processes, including attention^8–14 (see review by Paffen and Alais¹⁵). Because it provides information on the temporal dynamics of interocular suppression and binocular integration, binocular rivalry is widely used in both basic and clinical research.

Clinically, binocular rivalry has been used to find perceptual differences in populations with autism,¹⁶ bipolar disorder,¹⁷ or glaucoma,¹⁸ and the mixed percept in binocular rivalry has been shown to predict symptom severity of patients with attention-deficit/hyperactivity disorder.¹⁹ Mixed perception is one of the characteristics of binocular rivalry, which is a moment when incompatible images from both eyes are perceived simultaneously. It includes various perceptual experiences—namely, piecemeal^20,21 or superimposition^22,23—which shows that the visual system has different ways to integrate conflicting information between the eyes. Moreover, binocular rivalry serves as a sensory marker for interocular differences, such as in amblyopia,^24,25 or as a sensory dominance marker.^26,27 Lastly, binocular rivalry behavior has been associated with objective neuroscientific markers such as functional magnetic resonance imaging,²⁸ pupillometry,²⁹ or electrophysiological studies.^30–33

The current study was part of a large project that aimed to investigate the effect of attentional eye selection on binocular rivalry behavior and its electrophysiological correlates. The first aim of this study was to examine whether monocular attentional eye selection prolongs dominance of that same eye and how stimulus size affects this effect behaviorally when using a two-frequency tagging approach. Various stimulus sizes were used to examine how spatial scale influences the effectiveness of attentional eye selection. Larger stimuli may engage broader neural populations and influence receptive fields, potentially modulating how attention affects sensory dominance. The two-frequency tagging approach can enable the separation of each eye's visual evoked potential (VEP) during binocular rivalry,^13,31 which will be used for our next study to investigate the neural correlates of binocular rivalry. This two-frequency tagging approach was distinct from the static stimulus approach reported in previous investigations on selective attention.^9,34 Therefore, this study would also serve as a proof of concept of a suitable stimulus size for our next VEP study.

Studies have shown that attention affects sensory eye dominance; for example, attention can modulate overall dominance duration (i.e., the sum of the left and right eyes’ exclusive durations) or change the alternation rate.^10,34,35 Diverting attention away from the rival stimuli results in a reduction of binocular rivalry.¹³ Ooi and He⁹ found that directing attention to one image makes it less likely to be suppressed, whereas directing attention to the rival stimuli promotes it to reach visual awareness. They showed that the appearance of an attentional cue in one eye directs the perception toward the stimuli shown to that eye, suggesting an attentional effect on eye selection. This eye-specific selection by attention was further confirmed later by Zhang et al.,³⁶ who demonstrated that voluntary attention can modulate the processing of eye-specific visual information. Studies by Chong et al.¹² and Hancock and Andrews³⁷ demonstrated that both exogenous and endogenous attention prolong dominance duration. On the other hand, Meng and Tong¹⁰ reported that binocular rivalry is less susceptible to attentional bias, suggesting a more automatic process in binocular rivalry. It is noteworthy that these earlier experiments typically provided attention cues in a random or balanced manner across the two eyes, leaving the effect of monocular attention unaddressed. Therefore, this study introduced a critical modification from previous studies: rather than cueing randomly between the eyes, attention was deliberately directed to only one eye (i.e., the nonsighting eye in the current study) to investigate whether selective monocular attention can alter intrinsic eye dominance in binocular rivalry.

Our second aim was to explore whether attentional eye selection, by boosting the dominance of the selected eye, correspondingly reduces the proportion of mixed percepts. Therefore, in this study, two conditions (with and without attention) across stimulus sizes (1°, 2.5°, 5°, 10°, 15°) were tested to examine our aims.

Methods

Participants

In total of 12 observers with normal vision (6 females) between 22 and 60 years old (mean ± SD: 34 ± 13) participated in the study. The participants were recruited from the San Francisco Bay Area with a research advertisement. All participants had an eye examination by a pediatric ophthalmologist (author CH). Each eye had normal or corrected-to-normal vision (20/20 or better) measured with a Bailey-Lovie LogMAR chart and at least 40 arcsec stereoacuity estimated by using the random-dot stereo butterfly card (Stereo Optical, Chicago, Illinois). Their sighting and nonsighting eyes were determined using the hole-in-the-card test, in which 7 of 12 observers had a right-eye sighting dominance. The hole-in-the-card test was repeated at least twice at both the far distance (6 m) and the experimental distance (90 cm). All 12 participants were confirmed to have a consistent dominant eye at both distances. All participants except authors ZZ and CH were naive to the purpose of the study. The research protocol was approved by the Institutional Review Board of The Smith-Kettlewell Eye Research Institute and conformed to the tenets of the Declaration of Helsinki. Written informed consent was obtained before conducting the experiments.

Experimental Design

Stimuli were generated using MATLAB software (version 2019b; MathWorks, Natick, MA, USA) and the Psychophysics Toolbox 3,^38–40 run on a Mac computer. A pair of achromatic, orthogonal sinusoidal gratings (2 cycles per degree [cpd] spatial frequency) with equal contrast at 40% was presented on two matched 25-inch Sony OLED monitors (PVMA250 25-inch Professional OLED Production; Sony, Tokyo, Japan) viewed through a mirror stereoscope that was mounted on a chin rest at an effective distance of 90 cm. Each screen had a resolution of 2560 × 1440 pixels and was refreshed at 60 Hz. The luminance of the background was 50 cd/m². The gratings were contrast-reversed at 6 Hz for the vertical gratings presented to the sighting eye and at 7.5 Hz for the horizontal gratings presented to the nonsighting eye at 2 cpd, as shown in Figure 1. A spatial frequency of 2 cpd was selected because it elicits maximum contrast sensitivity in visual psychophysics⁷ and has been identified as optimal for VEP stimuli in previous electroencephalography studies.^8–10 This choice also aligns with our plan to use the same stimulus parameters in a subsequent VEP study, as stated in the Introduction. To minimize variability, we used a fixed spatial frequency across the stimulus sizes. The rival stimuli were varied in size at 1°, 2.5°, 5°, 10°, and 15° in random order. A central cross mark (0.2° at 80% contrast) was used as a fixation object.

Figure 1.

Stimuli and instruction. (A) Schematic experimental setup: a pair of 2 cpd orthogonal sinusoidal gratings at 40% contrast in each eye was presented to the two eyes viewing through a mirror stereoscope, with the horizontal gratings presented to the nonsighting eye. A contrast reversal of 7.5 Hz on the horizontal gratings and 6 Hz on the vertical gratings was employed. (B) Schematic of example percepts and the respective report instructions for perceptual exclusivity of the horizontal (press and hold key “1”) and vertical percepts (“2”), as well as all other mixed percepts (no key press). (C) Scheme of stimulus sizes used. Binocular contrasts and spatial frequencies were kept fixed while the sizes varied binocularly at 1°, 2.5°, 5°, 10°, and 15° in a random order.

Procedure

Participants sat in a chair and placed their heads onto a chin and forehead rest while viewing through the mirror stereoscope (Fig. 1A). Before beginning the experiment, the participants learned how to adjust the mirror stereoscope and to achieve fusion with the nonius line. This procedure was repeated during breaks between blocks to ensure that images from the two eyes were fused. Then, the participants had one or two practice blocks to experience binocular rivalry and to ensure they understood the instructions. The participants were instructed to fixate on a central mark while actively monitoring their own perceptual experience by reporting three perceptual categories through button presses: horizontal grating—press and hold the button “1,” as shown in Figure 1B; vertical grating—press and hold the button “2”; and mixed percepts, including piecemeal and superimposed perception—“Do not press any button.”

Two conditions, (1) without attention and (2) with attention directed to the horizontal gratings presented to the nonsighting eye, were tested for five stimulus sizes from all 12 observers, with five repetitions in each condition. Each trial lasted 42.3 seconds. The condition without attention served as the baseline, with observers viewing stimuli without specific attentional instructions. In the condition with attention, observers were instructed explicitly to direct their attention to the horizontal grating presented to one eye (observers were not aware of which eye saw the horizontal gratings). The two conditions were interleaved in a random order. Observers received short breaks (approximately 2–3 minutes) between blocks of trials to prevent fatigue.

Data Analysis

Raw data were stored as text files. Careful postprocessing steps were performed subsequently as described in the following. First, all responses within each trial that were shorter than 180 ms (see Skerswetat et al.⁴¹) were excluded to ensure that no data due to incomplete key presses or finger releases confounded the results. Second, as the last percept was artificially stopped at the end of each trial, the last percept was also removed from the main analysis. Next, the number of perceptual occurrences, perceptual durations, and relative proportions of each perceptual report (i.e., horizontal, vertical, or mixed percepts) were counted for each trial, condition, and participant.

Statistical Analysis

Excel (Microsoft Corp., Redmond, WA, USA) was used, along with the Real Statistics Resource Pack add-in program (Charles Zaiontz; http://www.real-statistics.com), for statistical analysis. Data outliers were identified using the 1.5 interquartile range rule, as shown in the figures. Therefore, nonparametric statistics (i.e., the Friedman test with Conover post hoc analysis and the Wilcoxon signed-rank test) were used in the study to consider the outliers. Correlation coefficients and significances were calculated using a two-tailed Spearman's ρ. To assess the variability of the observed effects in large samples, a bootstrapping procedure was implemented using MATLAB (version 2021b). Data from all subjects were resampled with replacement to generate 2000 bootstrap samples for each measure: perceptual occurrence, duration, and proportion. A fixed random seed was set before the bootstrapping process to ensure consistency, such that the same resampled subject pools were used across all three measures.

Results

Attentional Eye Selection Modulates Sensory Eye Dominance in Binocular Rivalry

Figure 2 plots the main results, which include data from the condition without attention (left column) and the condition with attention (middle column). Although statistical analysis in each stimulus size did not reveal a significant difference between the sighting and the nonsighting eyes (P > 0.05 across stimulus sizes), a numerical trend suggests a transition from the sighting eye dominance in the no-attention condition to attentional selected nonsighting eye dominance under the attention condition (Fig. 2H). A similar transition was also observed with a longer duration in the attentional selected nonsighting eye in Figure 2E, when compared to the sighting eye in Figure 2D. To better visualize this transition due to attentional eye selection, the differences between the two eyes were calculated in each condition for each size, as shown by the mean difference in Figures 2C, 2F, 2I, in which

$$\begin{array}{ccc}& & \mathrm{Red}:\left(withoutattention\right)\hfill \\ & & =\overline{SightingEye\left(without\mathit{attention}\right)-\overline{NonSightingEye\left(without\mathit{attention}\right)}}\hfill \end{array}$$

$$\begin{array}{ccc}& & \mathrm{Blue}:\mathit{Difference}\left(with\mathit{attention}\right)\hfill \\ & & =\overline{SightingEye\left(withattention\right)-\overline{NonSightingEye\left(withattention\right)}}\hfill \end{array}$$

Figure 2.

Comparison between the sighting and nonsighting eyes under conditions with and without attention from 12 observers. Left (A, D, G) and middle columns (B, E, H): boxplots of results from conditions without and with attention, respectively, in which data points (circles) that fall outside the whiskers are indicated as the outliers. Right column (C, F, I): difference means between the sighting and the nonsighting eyes in each condition. Error bars denote standard errors of the mean. The top part of the panel (pink area) corresponds to performance bias to the sighting eye; the bottom part of the panel (blue area) corresponds to performance bias to the nonsighting eye. Data revealed that sensory dominance favors the sighting eye (red bars) under the no-attention condition. However, when directing attention to the nonsighting eye, sensory dominance shifted from the sighting eye to the selected nonsighting eye (blue bars), particularly at stimulus sizes ranging from 2.5° to 10°, with the most pronounced effect observed at 5°. *P < 0.05, **P < 0.01, and ***P < 0.001, as determined by the Friedman test with Conover post hoc analysis (see details in Table 1).

As demonstrated in the right column of Figures 2C, 2F, 2I, the sighting eye (red bars) led the sensory dominance by exhibiting more frequent occurrences, longer dominance durations, and greater proportions than the nonsighting eye during binocular rivalry. When selectively directing attention to stimuli of the nonsighting eye (blue bars), the sensory eye dominance shifted from the sighting eye to the selected nonsighting eye, with the effect being most pronounced at a size of 5°. Such a sensory eye dominance shift was also confirmed by statistics at stimulus sizes 2.5°, 5°, and 10°, as indicated in Figures 2C, 2F, 2I. The relevant statistics are provided in Table 1.

Table 1.

Statistical Results for Figure 2

		Without Attention		With Attention		Comparison
Measure	Stimulus Size	Median	IQR/2	Median	IQR/2	t-Statistic	P Value
Occurrence difference (#)	1°	0.750	0.517	−0.250	0.7	1.003	0.317
	2.5°	0.300	0.688	−0.150	1.929	1.326	0.187
	5°	0.600	0.750	−0.458	1.433	2.115	0.036
	10°	0.750	0.763	−0.500	0.779	3.011	0.003
	15°	0.250	0.388	0.125	0.513	0.645	0.520
Median duration difference (s)	1°	−0.032	0.160	−0.335	0.188	1.722	0.088
	2.5°	0.035	0.102	−0.096	0.214	2.052	0.042
	5°	0.046	0.059	−0.304	0.294	3.151	0.002
	10°	0.102	0.100	0.020	0.261	1.905	0.059
	15°	0.029	0.248	0.061	0.173	1.136	0.258
Proportion difference (%)	1°	1.726	1.363	−2.999	7.941	1.905	0.059
	2.5°	3.474	4.061	−4.986	8.206	2.362	0.020
	5°	2.935	5.094	−7.678	9.478	3.963	0.000
	10°	5.298	4.257	−0.886	10.591	3.271	0.001
	15°	4.771	7.631	2.163	7.740	0.838	0.403

IQR, interquartile range.

Bolded values indicate statistical significance, with P <= 0.05.

Friedman-Conover test between the measured difference (sighting eye – nonsighting eye) under conditions with and without attention from 12 observers.

In the meantime, we also observed substantial individual differences, as shown in Figure 2. Although individual differences within and across subjects are one of the characteristics in binocular rivalry,^1,42 further analysis was performed to confirm the observations in Figure 2 that were not influenced by such individual differences from our small sample size (N = 12). We conducted bootstrapping with 2000 resampling iterations using data from 12 observers, as shown in Figure 3. Although the data still revealed large individual differences with a large sample (2000 resampling), as reflected in the long tails of the density distributions across both conditions (Fig. 3, three columns from left), the central tendency of the distributions still highlights our observations in Figure 2 from a small sample. For example, the transitions from the sighting eye (red) in the condition without attention (Fig. 3I) represent a greater proportion compared to the attentional selected eye in the condition with attention (Fig. 3J), which shows a greater proportion across all stimulus sizes, and these transitions are even clearer than those shown in Figures 2G, 2H. We also conducted bootstrapping for the difference between the two eyes in each condition, as shown in the third column of Figure 3. The right column of Figure 3 is the replot of the third column of Figure 3. Because the mean (black) and median (cyan) closely overlapped, only the mean with standard errors is shown for a better visualization and comparison with the results from nonbootstrapping data in Figure 2 (right column). Again, the shifts from the sighting eye, which is the sensory dominant eye under the no-attention condition, to the attentional selected nonsighting eye, which is the sensory dominant eye across all stimulus sizes, are substantial and highly significant, and these findings were confirmed by statistical tests (see Table 2). These results provided evidence that attentional eye selection can modulate sensory eye dominance.

Figure 3.

Distribution of variance in data obtained through bootstrapping individual observers’ data. Left column (A, E, I): results from the condition without attention. Second-to-left column (B, F, J): results from the condition with attention. Second-to-right column (C, G, K): bootstrapping of difference between the sighting and the nonsighting eyes. The y-axis shows density distributions of 2000 resampling iterations. Data illustrate the dynamics of binocular rivalry as a function of stimulus size in occurrence (top row), duration (middle row), and relative proportion (bottom row). Colors indicate the sighting eye (red) and the nonsighting eye (blue). Horizontal cyan and black bars represent the median and mean of data variance, respectively. Error bars show the standard deviation of the mean based on the bootstrapping procedure. Right column (D, H, L): replot of bootstrapping data from the second-to-right column with the difference mean for a comparison of Figure 2 (right column), in which the error bars denote the standard deviation of the mean. ***P < 0.001, as determined by the Friedman test with Conover post hoc analysis (see details in Table 2).

Table 2.

Statistical Results for Figure 3

		Without Attention		With Attention		Comparison
Measure	Stimulus Size	Median	IQR/2	Median	IQR/2	t-Statistic	P Value
Occurrence difference (#)	1°	0.655	0.131	−0.568	0.345	119.509	<0.001
	2.5°	0.125	0.334	0.929	0.263	95.7054	<0.001
	5°	0.408	0.314	−0.422	0.269	72.5097	<0.001
	10°	−1.139	0.412	0.1523	0.269	100.531	<0.001
	15°	0.948	0.274	−0.229	0.265	116.323	<0.001
Median duration difference (s)	1°	−0.006	0.051	−0.326	0.057	122.873	<0.001
	2.5°	0.062	0.031	−0.179	0.074	106.812	<0.001
	5°	0.083	0.024	−0.337	0.075	192.099	<0.001
	10°	0.113	0.042	−0.198	0.087	156.743	<0.001
	15°	0.153	0.065	−0.055	0.084	101.493	<0.001
Proportion difference (%)	1°	2.652	0.664	−6.447	2.299	114.936	<0.001
	2.5°	2.763	1.261	−8.096	3.021	152.862	<0.001
	5°	4.826	1.054	−10.121	3.039	249.264	<0.001
	10°	6.347	1.381	−7.013	2.984	223.879	<0.001
	15°	3.834	1.720	−2.854	2.743	87.050	<0.001

Bolded values indicate statistical significance.

Friedman-Conover test between the measured difference (sighting eye – nonsighting eye) under conditions with and without attention from bootstrapped data in Figures 3C, 3G, 3K.

Sighting Eye Dominance Effect during Binocular Rivalry

As noted above, sensory eye dominance was biased toward the sighting eye (red bars in the right column of Figs. 2 and 3) under the condition without attentional eye selection. An additional analysis was performed to test whether the effect was simply driven by an effect of “eye” rather than sighting eye dominance. In our cohort of 12 observers, 7 had the right eye as the sighting eye. Thus, the data were restructured to compare the left and right eyes (Fig. 4), rather than to compare the sighting and nonsighting eyes (Figs. 2, 3). Figure 4A plots the difference means between the left and the right eyes from all participants, showing no sensory dominance bias between the eyes, confirmed by statistical tests (Wilcoxon signed-rank test for one sample indicated P > 0.05 in all stimulus sizes). This observation, therefore, reinforces the earlier finding (Figs. 2, 3) that sensory eye dominance is specifically biased toward the sighting eye under the no-attention condition.

Figure 4.

Dynamics of binocular rivalry after regrouping the data with the performances from the left eye and the right eye. (A) Differences between the two eyes from 12 observers. (B) Distribution of variance obtained through bootstrapping (2000 resampling iterations) from individual observers. Horizontal black bars represent the mean of the data variance. Error bars show the standard deviation of the mean based on the bootstrapping procedure. Wilcoxon signed-rank test for one sample (data in A) and for two samples (data in B) revealed no significance in performance differences between the left and right eyes (P > 0.05 for all stimulus sizes).

Effect of Different Contrast-Reversal Rates on Perception during Binocular Rivalry

As noted earlier, the approach in the current study employed a different contrast-reversal rate in each eye, which was distinct from the static stimulus approach reported in previous investigations on selective attention.³⁴ The data, reanalyzed by comparing the left and right eyes (Fig. 4), also served as a test for whether the different contrast-reversal rate affects perception during binocular rivalry. In the current study, nearly half of the participants were left-eye dominant, while the other half were right-eye dominant. This indicates that the left and right eyes were exposed to different contrast-reversal rates (i.e., 6 Hz and 7.5 Hz). Consequently, the lack of significant differences between left- and right-eye performances provided evidence that different contrast-reversal rates in the two eyes did not induce differences in perception during binocular rivalry.

Stimulus Size Effect in Binocular Rivalry

As shown in Figure 3 (left two columns), stimulus size influenced the dynamics of binocular rivalry, producing an “n”-shaped pattern in perceptual alternation occurrence (Figs. 3A, 3B) and a “u”-shaped pattern in dominance duration (Figs. 3E, 3F) across stimulus sizes ranging from 1° to 15°. The 5° stimulus elicited the most frequent perceptual alternations and the shortest dominance durations. This inverse relationship appeared consistent across both sighting and nonsighting eyes (Fig. 3, left two columns), as well as left and right eyes (Fig. 4B, left two panels). When attention was directed to stimuli presented to the nonsighting eye, the strongest attentional effect, marked by a greater transition from the sighting eye to the selected nonsighting eye, was observed at the 5° stimulus size (Fig. 2, right column).

Mixed Perception

The attention effect on mixed perception across stimulus sizes, although not statistically significant (P > 0.05 for all stimulus sizes), is evident, as shown in Figures 5A, 5B. Given the small sample size, a bootstrap analysis was conducted on the individual data, as shown in Figure 5C. The data revealed large individual differences, especially at the smallest stimulus size (1°), in perceptual dominance duration (Fig. 5, middle column), reflected in the outliers (Fig. 5A, middle panel) and the long tails of the density distributions (Fig. 5C, middle panel) across both conditions. Although the data revealed no significant differences in mixed perception between the attention and no-attention conditions from bootstrapped large sample data in each measure (Fig. 5C, P > 0.05 for all stimulus sizes), a trend of decreasing proportion with increasing stimulus size was observed, as shown in Figures 5A, 5C (right panels). The correlation in Figure 5A (right panel) was significant (Spearman's ρ = −0.99, P < 0.001) by the average of the data from the conditions with and without attention. This trend appears to be primarily driven by the high proportion observed at the smallest stimulus size (1°).

Figure 5.

Comparison of mixed perception between no-attention and attention conditions as a function of stimulus size. (A) Boxplots of mixed perception, in which data points (circles) that fall outside the whiskers indicate the outliers. Colors denote test conditions. (B) Differences between attention and no-attention conditions from 12 observers. Error bars denote the standard error of the mean. (C) Distribution of variance obtained through bootstrapping (2000 resampling iterations) from individual data. Left column: mixed perception occurrence. Middle column: mixed perception duration. Right column: mixed perception proportion. Colors denote the condition with and without attention. Horizontal black bars represent the mean of the data variance. Error bars show the standard deviation of the mean based on the bootstrapping procedure. Wilcoxon signed-rank test for two samples (A, C) and one sample (B) revealed no significance in each measure between the no-attention and attention conditions in mixed perception across the stimulus sizes (P > 0.05).

Discussion

Effect of Attentional Eye Selection in Binocular Rivalry

The first aim of this study was to determine whether selective monocular attention can alter intrinsic ocular dominance. Although previous studies have reported mixed findings regarding the relationship between sighting eye dominance (measured by the hole-in-the-card test) and sensory eye dominance (measured by binocular rivalry), some studies found no correlation,^43–45 while others reported significant correlations.^2,46,47 Our results showed that in the no-attention condition, the sighting eye leads the sensory dominant eye, evidenced by a higher occurrence rate, longer dominance duration, and larger dominance proportion for the sighting eye (see the pink areas of Figure 2, right column, and Figure 3, right two columns). This no-attention condition allowed us to use sighting eye dominance as the intrinsic eye dominance baseline to evaluate the effects of attentional eye selection. Crucially, directing attention to the nonsighting eye reversed the dominance pattern: the previously less perceptual proportion of the nonsighting eye in the condition without attention (Fig. 2G, blue bars) became a more perceptual proportion in the condition with attention (Fig. 2H, blue bars). This perceptual proportion transition between the eyes affected by selective attention is even more clearly demonstrated in a comparison of the difference between the eyes (Fig. 2I, red bars in the condition without attention versus blue bars in the condition with attention). These findings demonstrated that attentional eye selection can modulate sensory eye dominance during binocular rivalry.

To ensure that the observed shift in sensory eye dominance was not due to individual variability or a simple left-versus-right eye preference, we conducted two control analyses. First, a 2000-iteration bootstrap confirmed the reliability of the sensory eye dominance reversal across observers (Fig. 3). Second, regrouping the data by left versus right eye eliminated the effect, ruling out fixed eye-of-presentation biases (Fig. 4). Together, these control analyses support the conclusion that the observed change in sensory eye dominance was driven by eye-selective attention rather than by inherent eye-based factors.

While these findings demonstrate that attentional eye selection can modulate sensory eye dominance during rivalry, there are several limitations. One limitation is the potential contribution of response bias. One might ask whether the observed sensory eye dominance reversal could reflect a response bias rather than a genuine perceptual change. Two aspects of the data argue against a purely response-based account. First, the dominance reversal was accompanied by a systematic increase in the mean dominance duration of the attended eye and a corresponding decrease for the unattended eye, a pattern not easily explained by a mere shift in response criterion. Second, the proportion of mixed percepts did not increase in the attention condition (Figs. 5A, 5C); if participants were guessing during ambiguous moments, an increase in mixed reports would be expected. Additionally, a fully implicit attention manipulation, such as rapid exogenous cueing or eye-unspecific task demands, was not feasible in the current design, as the cued stimulus was fixed to one eye. Future studies could address this limitation by combining our two-frequency tagging approach with objective neural markers (e.g., steady-state visual evoked potential (SSVEP) gain) to better distinguish between response biases and true perceptual shifts. While we recognize response bias as a potential limitation, we consider it unlikely to fully explain the robust reversal in eye dominance observed here.

One possible neural mechanism underlying attentional eye selection in binocular rivalry is that top-down influences modulate early visual processing in a manner that mimics the effect of increasing contrast to the attended stimulus.^48,49 This attentional modulation is consistent with Levelt's proposition I, which states that increasing stimulus strength for one eye enhances the perceptual dominance of that eye's stimulus. Moreover, the reversed dominance pattern observed when attention was directed to the nonsighting eye (Fig. 2G versus Fig. 2H, blue bars) suggests that attentional eye selection influences binocular rivalry dynamics for both eyes. Specifically, attending to the nonsighting eye increased its perceptual proportion while decreasing that of the sighting eye. These findings are consistent with prior attention literature showing that attention enhances the processing of attended targets (i.e., the nonsighting eye in our case) while suppressing irrelevant distractors (i.e., the sighting eye).^50–52 Taken together, these results suggest that directing attention to one of the rival images can bias perception in favor of that stimulus, increasing both its likelihood of being perceived and its dominance duration.^12,53 These findings support the notion that attention contributes to resolving the competition between rival images during binocular rivalry.^{9,10,12–14,54}

Stimulus Size Effect in Attentional Eye Selection during Binocular Rivalry

While the preceding discussion clarified the role of attention in influencing binocular rivalry, the stimulus size could be another important factor driving this process. Therefore, the second aim was to determine whether stimulus size affects the magnitude of this flip. The perceptual mean duration exhibited a U-shaped profile, with the shortest and most frequent alternations occurring at 5°. This finding, particularly within the 1° to 5° stimulus size range, is consistent with previous studies. Kang⁵⁵ reported that a larger stimulus (0.8° × 3.6°) produced a shorter dominance duration compared to a size of 0.8°, although only two sizes were tested. Similarly, Law et al.⁵⁶ found that a small stimulus of 0.5° resulted in slower alternation rates and longer durations compared to 1° and 1.5°. The most relevant comparison is from O'Shea et al.,⁷ who examined stimulus fields ranging from 0.5° to 8° and found the highest proportion of exclusive visibility at sizes between 2° and 4° in three participants. Our findings extend this work by testing a broader size range (1°–15°) in a larger sample (N = 12), revealing a peak in perceptual dominance between 2.5° and 10° (Fig. 2, bottom row). Notably, when attention was directed to the nonsighting eye, the strongest attentional effect occurred at 5°, a finding not previously reported in the literature.

Regarding mixed perception, previous studies on binocular rivalry and stimulus size have reported that as stimulus size increases, the proportion of exclusive visibility decreases while mixed perception increases.^7,41,57 In contrast, our findings showed the opposite pattern (Figs. 5A, 5C). Prior work using static grating stimuli indicated that increasing the size of dichoptic stimuli beyond a 4° diameter reduced exclusive visibility and enhanced mixed perception.⁷ In the present study, we employed a two-frequency tagging approach,³¹ where each eye received a stimulus with a distinct contrast-reversal rate. This setup was designed as a behavioral proof of concept for our future electrophysiological studies on eye-selective attention. For example, our future electrophysiological studies aim to investigate how the brain suppresses one eye's stimuli (e.g., vertical grating) when the other eye sees horizontal grating, and how attentional eye selection influences neural processing during binocular rivalry. Across stimulus sizes, attentional eye selection did not significantly influence the proportion of mixed perception. Importantly, the use of different contrast-reversal rates did not itself bias perceptual outcomes during rivalry, as confirmed by additional data analyses (Fig. 4). However, it remains unclear which specific perceptual categories (for more details, see Skerswetat and Bex⁵⁸) contributed to this outcome. Further studies may investigate the categories of mixed percepts and their contribution when increasing stimulus sizes. For example, previous studies suggest that piecemeal percepts may reflect local zones of rivalry,⁵⁹ whereas superimposition may indicate temporal binocular fusion.²²

Clinical Implications

The findings of the present study, particularly the demonstration that attentional eye selection can modulate sensory eye dominance, have important implications for clinical practice in ophthalmic care, highlighting the role of attention in the rehabilitation of neuro-ophthalmic disorders.⁶⁰ For example, directing attention to the amblyopic eye or incorporating attention-demanding tasks in a dichoptic training paradigm may offer an effective approach for amblyopia treatment. Several training studies support this notion.^27,61 For instance, Ooi et al.²⁷ employed a binocular rivalry-based push–pull protocol involving visual cueing to the amblyopic eye, resulting in improvements in stereoacuity and a reduction in interocular suppression. These behavioral improvements were hypothesized to result from a recalibration of the excitatory/inhibitory balance between the eyes.⁶² Hou and Nicholas⁶¹ used high-attention-demand tasks for the amblyopic eye in a dichoptic setting as a training tool and reported a reduction in interocular suppression that correlated with visual acuity improvement in the amblyopic eye. These successful training results were even observed in adult amblyopes over 50 or 60 years old. Additionally, a prior study has shown that attentional eye selection can shift sensory eye dominance following short-term monocular treatment.⁵⁴

Although these results suggest that incorporating attention-based tasks targeting the amblyopic eye to modulate sensory eye dominance may serve as an effective training strategy, a limitation of the current study is that all participants were adults. Given that amblyopia is predominantly a pediatric condition and that attentional selection mechanisms may differ between adults and children due to developmental factors, the applicability of these findings to the pediatric population remains uncertain. Future studies should investigate whether attentional modulation of eye dominance exhibits a similar pattern in typically developing children.

Conclusions

This study demonstrated that the sighting eye leads sensory eye dominance during binocular rivalry, as indicated by more frequent perceptual occurrences, longer dominance durations, and greater perceptual proportions compared to the nonsighting eye. However, when attention was selectively directed to stimuli in the nonsighting eye, sensory eye dominance shifted in favor of that eye. Stimulus size also influenced the dynamics of binocular rivalry and sensory eye dominance under attentional eye selection. Notably, attentional eye selection did not influence the proportion of mixed perception, regardless of stimulus size. These findings suggest that attentional eye selection can modulate sensory eye dominance. These results may have clinical implications for understanding and managing visual disorders influenced by interocular sensory imbalances, such as amblyopia.