Increases versus decreases: Asymmetric effects of contrast changes during binocular rivalry modulated by awareness of perceptual switch

Abstract

The human visual system prioritizes dynamic stimuli, which attract attention and more readily break suppression to reach perceptual awareness. Here, we investigated whether dynamic changes in contrast—either increasing or decreasing—are equally effective in facilitating the breakthrough of suppressed stimuli during binocular rivalry. In Experiment 1a, we found that contrast increases led to significantly faster breakthroughs into perceptual dominance compared with decreases. Notably, increases accelerated breakthrough relative to the unchanged baseline, whereas decreases delayed it. Experiments 1b and 1c replicated the results of Experiment 1a using, respectively, a briefer contrast change (10 ms instead of 100 ms) and partial breakthrough reports, confirming a robust asymmetry in the processing of suppressed stimuli between increases and decreases. In Experiment 2a, random dots moving in different random directions were presented dichoptically, making interocular conflict imperceptible and unreportable. We found that any change in intensity in such rivalry settings—regardless of increase or decrease—promoted perceptual dominance. By introducing motion stimuli into the Experiment 1 paradigm, Experiment 2b demonstrated that the divergence between Experiments 1 and 2 was not due to low-level stimulus differences. Taken together, our results reveal an asymmetric effect of contrast changes during binocular rivalry. This finding highlights the interplay between subliminal sensory processing of contrast changes and conscious awareness, shedding light on developing theoretical models of binocular rivalry.

Introduction

The human visual system prioritizes novel or changing stimuli over static inputs, facilitating rapid responses to potential threats or opportunities in dynamic environments (Candan Şimşek, Karaca, Kırmızı, & Ekiz, 2023; Franconeri & Simons, 2003; Rensink, 2002; Schomaker & Meeter, 2015). This adaptive preference exhibits a directional asymmetry: Visual stimuli that suddenly appear or increase in intensity capture attention more effectively than those that disappear or decrease (Cole, Kentridge, & Heywood, 2004; Yantis & Jonides, 1984). Similar biases are observed in auditory perception, where approaching or increasing intensity sounds are perceived as more salient and urgent compared with receding or decreasing intensity sounds (Ignatiadis, Baier, Tóth, & Baumgartner, 2021; Neuhoff, 1998; Neuhoff, 2001). These asymmetries are thought to reflect an intrinsic bias that promotes survival by prioritizing responses to potential threats or rewards.

Recent studies suggest that the visual system continues to process dynamic changes even without awareness. Binocular rivalry—in which conflicting images presented separately to each eye compete continuously for perceptual dominance—provides an ideal paradigm to examine such unconscious perceptual mechanisms (Blake, 2022; Blake & Logothetis, 2002; Wheatstone, 1838). Studies have found that a sudden change in a suppressed stimulus, such as a transient flash, can trigger an immediate breakthrough into awareness (Lee, Blake, & Heeger, 2005; Lin & He, 2009). However, it remains unknown whether this unconscious processing differentiates between the direction of change in stimulus intensity, as conscious perception does. Furthermore, because top–down factors such as attention and expectation can bias which stimuli gain dominance during rivalry (Gayet, Paffen, & Van der Stigchel, 2013; Gayet, Van Maanen, Heilbron, Paffen, & Van der Stigchel, 2016; Meng & Tong, 2004), it is critical to ask whether the effect of sensory intensity change depends on higher-level modulation. Clarifying these questions will help elucidate fundamental properties of subconscious visual processing, advancing our understanding of how intrinsic sensory biases shape perceptual awareness.

Here, we investigated how abrupt increases versus decreases in stimulus contrast affect perceptual dominance during binocular rivalry. Experiment 1 directly tested this question by comparing breakthrough times across increasing, decreasing, and static conditions. Experiment 2 utilized an unreportable rivalry paradigm adapted from Brascamp, Blake, and Knapen (2015). This method is designed to minimize contributions from top-down executive processes, such as report-related attention and expectation, which are known to modulate rivalry (e.g., Chong, Tadin, & Blake, 2005; Meng & Tong, 2004). This allows for a cleaner assessment of bottom–up sensory biases in perceptual selection when a change is applied to an unreportable stimulus.

Experiment 1. Asymmetric effects of contrast changes in suppressed rivalry stimuli

Experiment 1a

Methods

Participants

Eleven healthy individuals (age range, 18–23 years; 10 females) with normal or corrected-to-normal visual acuity participated in Experiment 1a. The sample size for this initial exploratory experiment (n = 11) was determined in line with prior studies in the field (e.g., Blake, Tadin, Sobel, Raissian, & Chong, 2006; Shimizu & Kimura, 2023). A post hoc power analysis confirmed that the sample was adequately powered to detect the large effect size observed. None of the participants, except the author, was aware of the purpose of the experiment. The study adhered to the tenets of the Declaration of Helsinki and received approval from the Human Research Ethics Board at South China Normal University. Written informed consent was obtained from each participant, and they were compensated for their involvement. The same ethical standards were followed in all subsequent experiments.

Apparatus

To display the stimuli, we used a gamma-corrected 24-inch VPixx CRT screen (VPixx Technologies, Saint-Bruno, QC Canada), with a 100-Hz refresh rate, 1920 × 1080 resolution, and peak brightness of 95 cd/m². Gamma correction was performed following Psychtoolbox-3 guidelines (Brainard, 1997). Luminance was measured using a CS-150 Luminance and Color Meter (Konica Minolta, Tokyo, Japan). In a darkened room, participants viewed the screen through a mirrored stereoscope, their heads stabilized by a chin rest at a viewing distance of 60 cm. Experimental programs were executed on a Dell PC (Dell Technologies, Round Rock, TX) running MATLAB (MathWorks, Natick, MA). This setup was consistently employed in all experiments conducted in this study.

Stimuli

All stimuli were generated using MATLAB with Psychtoolbox-3 (Brainard, 1997) and presented on a gray background. A red fixation cross (diameter, 0.53°) was displayed to each eye to maintain gaze. The rivalry stimuli consisted of two diagonal sinusoidal gratings (diameter, 1.5°; spatial frequency, 5 cycles per degree), oriented at 45° and 135°. The contrast of each grating was defined using the Michelson contrast formula:

$$C_m=\frac{L_{max}-L_{min}}{L_{max}+L_{min}}$$

where L_max and L_min denote the maximum and minimum luminance, respectively (Michelson, 1927). To minimize the influence of static features on the rivalry process, the phase of each grating was updated every 10 ms, creating a constant motion of approximately 1.67 cycles per second. This manipulation discourages locking of gaze or attention onto any specific part of the stimulus, thus preventing potential response biases and allowing for more typical perceptual alternations. In both the increasing and decreasing conditions, the contrast of the grating of the suppressed eye was adjusted—either increased from 0.4 to 0.8 or decreased from 0.4 to 0.2—within 100 ms, following a logarithmic scaling (Stevens, 1961). To aid binocular alignment, a checkerboard frame was presented around the rivalry stimulus. This frame consisted of two square regions: an inner square measuring 2.8° × 2.8° and an outer square measuring 3.2° × 3.2°, with the area between them filled with a black-and-white checkerboard pattern. After each trial, a masking stimulus composed of superimposed gratings in multiple orientations was displayed to eliminate any lingering afterimages.

Design and procedure

This experiment used a within-subjects design. The independent variable was the type of contrast manipulation, with four levels: increase, decrease, static high-contrast baseline, and static low-contrast baseline. The dependent variable was the breakthrough latency. Each trial followed a fixed sequence: fixation, rivalry, contrast manipulation, perceptual report, and masking (Figure 1). At the start of each trial, participants fixated centrally. They were instructed to first take a moment to let the rivalry process stabilize and thereafter to press the left or right arrow key to indicate which grating was fully dominant. To ensure the quality of the report, they were explicitly instructed to respond only when one grating was perceived clearly and the other was completely invisible. All participants completed a practice session (24 trials) to familiarize themselves with this criterion. Each trial was assigned a target eye (the eye that would receive the contrast change) and a non-target eye. The contrast change was triggered immediately by any keypress reporting the non-target eye as dominant. This ensured that the breakthrough latency measurement always began at the onset of the suppression period of the target eye, regardless of the initial reported percept. Participants then continued monitoring and pressed the corresponding key when the manipulated stimulus achieved full dominance. The interval between the trigger and this report was recorded as the breakthrough latency. If no switch was reported within 20 seconds, the trial terminated automatically and was excluded from analysis. At the end of each trial, a 2-second mask was presented to minimize visual after-effects.

Figure 1.

Schematic of the procedure in Experiment 1. Each trial consisted of fixation, continuous rivalry, contrast manipulation, perceptual report, and masking. When participants pressed a key to report dominance of the preassigned non-target eye, a brief contrast change was introduced to the grating presented to the suppressed eye. Participants then pressed a key to report breakthrough of the manipulated stimulus. Blue and orange boxes illustrate contrast increase and contrast decrease conditions, respectively.

In addition to the increasing and decreasing conditions, baseline conditions with unchanged contrast were included for comparison. To control for differences in suppression time related to contrast magnitude, each change condition was paired with a corresponding baseline condition where the contrast of the grating of the target eye remained constant, matching the final contrast level of its associated change condition (e.g., the baseline for the increase condition was a static 0.8 contrast). All four conditions were presented in a randomized, interleaved order. In baseline trials, the procedure was identical except that no contrast change occurred; we measured the suppression duration from the report of non-target eye dominance until the natural breakthrough of the static target stimulus. The average of these single-trial latencies served as the comparison for its corresponding change condition.

Data analysis

All reported breakthrough times fell within this 20-second limit; trials were discarded only if they were identified as outliers (exceeding 3 SD from a participant's mean for that condition). This procedure retained approximately 97.64% of the trials for statistical analysis. The normality of the cleaned data was confirmed using the Shapiro–Wilk test (Shapiro & Wilk, 1965). Paired-sample t-tests were conducted separately for the increase versus decrease conditions, the increase versus static high-contrast baseline, and the decrease versus static low-contrast baseline.

Results

Stimuli with increasing contrast broke through suppression (1.676 ± 0.556 seconds) significantly faster than those with decreasing contrast (4.283 ± 1.405 seconds): t(10) = −6.339; p < 0.001; Cohen's d = −1.911; 95% confidence interval (CI), −2.91 to −0.88. This effect was robust across participants: As shown in Figure 2A, all 11 individuals exhibited this pattern, with shorter breakthrough times in the increasing condition. Figure 2B further illustrates the distributions of reaction times across participants, showing that increasing stimuli tended to elicit faster and more clustered breakthroughs (mostly within 2 seconds), whereas decreasing stimuli led to more variable and delayed responses.

Figure 2.

Breakthrough times under different contrast change conditions. (A) Breakthrough times in the increasing and decreasing contrast conditions. Each point represents an individual participant. Bars indicate group means, and error bars represent ±1 SEM. (B) Frequency distributions of breakthrough times across all participants for increasing (blue) and decreasing (orange) conditions. The histogram for author CH is labeled Sub3. (C) Comparison between dynamic contrast-change conditions and static baselines matched in final contrast. **p < 0.01, ***p < 0.001.

Compared to unchanged baseline conditions with matched final contrast, increasing stimuli produced significantly faster breakthroughs (1.676 ± 0.556 seconds vs. 1.979 ± 0.654 seconds): t(10) = −5.387; p < 0.001; Cohen's d = −1.624; 95% CI, −2.53 to −0.69. In contrast, decreasing stimuli were significantly slower than their static counterparts (4.283 ± 1.405 seconds vs. 3.985 ± 1.325 seconds): t(10) = 4.001; p = 0.003; Cohen's d = 1.206; 95% CI, 0.40 to 1.98. These results indicate that it is not only the final intensity but also the change in contrast itself that critically shapes perceptual competition. Moreover, the brain appears to process increases and decreases in contrast differently, suggesting an asymmetry in how dynamic visual signals are resolved during rivalry.

Experiments 1b and 1c

Experiment 1a revealed a striking asymmetry, suggesting that the direction of contrast change affects unconscious visual processing. However, two aspects of the original design may have contributed to this effect. First, the change duration (100 ms) may have been too slow to fully prevent conscious detection. Second, participants were instructed to respond only after fully perceiving the grating, raising the possibility that partial percepts—especially for decreasing stimuli—could have altered the rivalry dynamics. To test the robustness of this asymmetry, we conducted two supplementary experiments.

In Experiment 1b, we reduced the duration of contrast change to a single frame (10 ms) to further minimize conscious awareness of the change. Seven participants (age range, 20–22 years; four females) completed the task. The sample sizes for this and all subsequent experiments were determined by a single a priori power analysis (G*Power 3.1.9.7) (Faul, Erdfelder, Lang, & Buchner, 2007). Based on the large effect size observed in Experiment 1a (Cohen's d = −1.91), this analysis indicated that a sample of n = 5 would achieve 95% power. Our sample sizes (n = 7 or 8) therefore provide sufficient statistical power. Breakthroughs were significantly faster for increasing stimuli (1.361 ± 0.596 seconds) than for decreasing stimuli (4.174 ± 1.590 seconds): t(6) = −5.159; p = 0.002; Cohen's d = –1.950; 95% CI, −3.23 to −0.62. Compared to static conditions matched in final contrast, increasing stimuli were also faster than high-contrast static stimuli (1.361 ± 0.596 seconds vs. 1.632 ± 0.552 seconds): t(6) = −2.892; p = 0.028; Cohen's d = −1.093; 95% CI, −2.02 to −0.11. Decreasing stimuli, however, were slower than low-contrast static stimuli (4.174 ± 1.590 seconds vs. 3.671 ± 1.626 seconds): t(6) = 5.128; p = 0.002; Cohen's d = 1.938; 95% CI, 0.62 to 3.22 (Figure 3A).

Figure 3.

Supplementary experiments testing the robustness of contrast-change asymmetry. (A) Experiment 1b: Mean time to full breakthrough for increasing and decreasing contrast stimuli and their respective static baselines (higher, lower). Contrast changes occurred within a single frame (10 ms). (B) Experiment 1c: Mean time to partial breakthrough using a continuous keypress-release method. Bars indicate means; lines connect individual participants; error bars represent ±1 SEM. *p < 0.05, **p < 0.01.

In Experiment 1c, we modified the task to measure partial rather than full breakthroughs, providing a more sensitive measure of suppression duration. The contrast change was triggered identically to Experiment 1a; however, unlike the discrete-report method in Experiment 1a, participants here continuously reported dominance by pressing and holding a key. They were instructed to release it as soon as the percept became mixed or unstable, and this release time was recorded as the partial breakthrough latency. This continuous-response method captured finer perceptual transitions. Data from eight participants (age range, 19–23 years; six females) showed the same pattern: increasing stimuli broke through faster than decreasing ones (1.044 ± 0.525 seconds vs. 2.747 ± 1.766 seconds): t(7) = −3.572; p = 0.009; Cohen's d = −1.263; 95% CI, −2.19 to −0.29. The increasing condition also outpaced the static high-contrast baseline (1.044 ± 0.525 seconds vs. 1.268 ± 0.643 seconds): t(7) = −3.437; p = 0.011; Cohen's d = −1.215; 95% CI, −2.13 to −0.26. However, decreasing stimuli lagged behind the static low-contrast condition (2.747 ± 1.766 seconds vs. 2.199 ± 1.160 seconds): t(7) = 2.360; p = 0.050; Cohen's d = 0.835; 95% CI, 0.00 to 1.63 (Figure 3B). These results confirm the robustness of the effect across different temporal parameters and response criteria, highlighting that the direction of contrast change to a suppressed stimulus—not just final intensity—modulates perceptual dominance asymmetrically.

Experiment 2. Directional biases in unreportable binocular rivalry

Experiment 2a

Methods

Participants

Seven participants took part in Experiment 2a (age range, 18–24 years; five females). All had normal or corrected-to-normal vision and were recruited independently from those in Experiment 1.

Stimuli

The competing stimuli consisted of moving dots (radius = 0.08°; density = 138 dots per degree; speed = 5.7°/s) within an annular aperture (inner radius = 0.15°, outer radius = 1.25°). When a dot reached the aperture boundary, a new dot was symmetrically regenerated on the opposite side to maintain constant density. A rectangular checkerboard frame (2.8° and 3.2° square) surrounded the stimuli to aid binocular alignment. The central fixation point was 0.13°, smaller than the inner aperture radius. Its color was red (RGB: 255, 0, 0) during the invisible rivalry periods and green (RGB: 0, 255, 0) during the short binocular rivalry (BR) periods. The background was gray (RGB: 128, 128, 128). The contrast of the competing stimuli followed Weber contrast, defined as

$$C=\frac{I-I_b}{I}$$

where I and I_b represent the dot luminance and background luminance, respectively (Peli, 1990). The contrast parameters were selected with reference to Experiment 1.

Design and procedure

The experiment was adapted from the unreportable rivalry paradigm developed by Brascamp et al. (2015), which consists of an initial phase of invisible binocular rivalry followed by a brief phase allowing perceptual report (Figure 4). In the present study, we inserted a brief contrast-change manipulation between these two stages. During the invisible rivalry periods, 40% of the dots (signal dots) moved coherently in a designated signal direction but the remaining 60% (noise dots) moved randomly. Every 300 ms, dot directions were updated: Signal dots were reassigned a new coherent direction, and noise dots continued to move randomly. In each eye, the motion of the signal dot directions differed by at least 90°, creating interocular competition. The duration of each invisible rivalry period was randomly selected from four possible intervals (1.8, 2.4, 2.9, and 3.3 seconds), during which participants simply viewed the display without making any response. Although binocular rivalry occurred during this phase, participants could not consciously report the specific motion direction.

Figure 4.

Schematic of the procedure in Experiment 2a. Each trial began with a fixation period (1.5 seconds), followed by an invisible rivalry period (1.8–3.3 seconds) during which 40% of dots moved coherently (signal) and 60% moved randomly (noise). After a brief 100-ms contrast change, a short BR period (0.6 second) presented coherent motion in opposite directions to each eye. Participants then indicated whether the motion was perceived as leftward or rightward. Note that the actual stimulus contained a much higher dot density than shown here for clarity.

Following the invisible rivalry period, a brief contrast change was introduced: Contrast was scaled in multiplicative steps every 10 ms over a 100-ms interval while the dot motion was held constant. During the short BR period, all dots within each eye moved uniformly in a single direction for exactly 600 ms. In one eye, motion directions were centered within a 135° range around the leftward direction; in the other eye, it was centered within a 135° range around the rightward direction. Importantly, the contrast of the stimuli in the two eyes was identical during this period; the short BR phase served solely to reveal the dominant percept resulting from the preceding invisible rivalry period. At the end of each short BR period, participants indicated the perceived direction of motion (left or right) using a two-alternative forced-choice task.

Experiment 2 consisted of 210 trials. Four main conditions were tested, 2 contrast changes (increase/decrease) × 2 eyes (left/right), with 40 trials per condition. An additional 40 trials served as a baseline with no contrast changes during the critical 100-ms phase. Ten catch trials, in which the dots for both eyes moved identically during the short BR periods, were presented to verify participants’ attention.

Data analysis

We quantified perceptual bias at the level of individual eyes using the proportion of percepts consistent with contrast manipulation (PPCM). For each eye, PPCM was defined as the proportion of trials in which the perceived motion direction matched that of the eye assigned to a given contrast manipulation (increase, decrease, or no change). In total, data from 14 eyes (7 participants × 2 eyes) were analyzed. All participants achieved 100% accuracy on catch trials, confirming compliance with task instructions.

Results

Stimuli with increasing contrast were significantly more likely to dominate perception (PPCM = 0.685 ± 0.158) than those with decreasing contrast (0.557 ± 0.127): t(13) = 4.408; p = 0.001; Cohen's d = 1.178; 95% CI, 0.48 to 1.85 (Figure 5). Thus, the greater efficacy of a contrast increase over a decrease persists even when the interocular conflict is unreportable. Compared with the static baseline (0.500 ± 0.096), both types of contrast change increased the likelihood of dominance. The increase condition led to significantly more consistent percepts than the unchanged condition: t(13) = 6.144; p < 0.001; Cohen's d = 1.642; 95% CI, 0.81, 2.44. Even decreasing stimuli showed a modest but significant advantage over baseline: t(13) = 2.541; p = 0.025;Cohen's d = 0.679; 95% CI, 0.09 to 1.25. These findings suggest that a transient contrast change—regardless of direction—can bias perception in favor of the manipulated eye, even when the change is applied to an unreportable stimulus.

Figure 5.

Proportion of percepts consistent with contrast manipulation (PPCM) across conditions in Experiment 2a. Each bar represents the mean PPCM across eyes for increasing, decreasing, and unchanged conditions. Dots represent individual eyes (n = 14). Higher PPCM indicates a stronger perceptual bias toward the eye that underwent contrast manipulation. Error bars show ±1 SEM. *p < 0.05, ***p < 0.001.

Experiment 2b

Experiment 1 showed that contrast increases facilitated perceptual dominance, whereas contrast decreases delayed it. In Experiment 2a, however, both contrast increases and decreases appeared to facilitate dominance. To clarify whether this discrepancy was due to low-level stimulus differences, Experiment 2b was designed to directly replicate the procedure of Experiment 1a using moving dots. The stimuli were adapted from the short BR period of Experiment 2a, such that each eye viewed dots moving with 100% coherence in one of two orthogonal directions (e.g., leftward vs. rightward). This induced a standard, reportable state of binocular rivalry, allowing us to measure the breakthrough latency after a contrast change was applied to the suppressed motion stimulus. If the results of Experiment 2b replicated the pattern observed in Experiment 1, it would suggest that the effects observed in Experiment 2a were more likely due to a lack of conscious differentiation between competing stimuli, rather than stimulus complexity.

Eight participants (age range, 19–23 years; four females) completed the task. Consistent with previous experiments, increasing stimuli broke through suppression (1.226 ± 0.439 seconds) significantly faster than decreasing stimuli (3.015 ± 0.821 seconds): t(7) = −7.469; p < 0.001; Cohen's d = −2.641; 95% CI, −4.14 to −1.11. The increase condition was also faster than its high-contrast static baseline (1.226 ± 0.439 seconds vs. 1.485 ± 0.518 seconds): t(7) = −4.767; p = 0.002; Cohen's d = –1.685; 95% CI, −2.77 to −0.56. Conversely, decreasing stimuli were slower to break through than their static low-contrast counterparts (3.015 ± 0.821 seconds vs. 2.808 ± 0.814 seconds): t(7) = 3.069; p = 0.018; Cohen's d = 1.085; 95% CI, 0.17 to 1.95 (Figure 6). These results support the interpretation that the results observed in Experiment 2a were not due to the complexity of the dot motion stimuli but were more likely related to the participants’ inability to consciously differentiate the competing stimuli during rivalry.

Figure 6.

Replication of the asymmetry using random dot stimuli under full-report instructions (Experiment 2b). Breakthrough times for increasing and decreasing contrast conditions are shown alongside their corresponding static baselines. *p < 0.05, **p < 0.01.

Discussion

This study systematically investigated how subconscious changes in stimulus intensity influence perceptual selection during binocular rivalry. Across all experiments, we consistently observed a directional asymmetry: Increases in stimulus intensity facilitated perceptual dominance more strongly than equivalent decreases. Importantly, unlike increases, which consistently accelerated breakthrough relative to the unchanged baseline, the perceptual outcome of intensity decreases varied with the awareness of interocular conflict. Specifically, under conscious rivalry conditions (Experiments 1a, 1b, 1c, and 2b), intensity decreases delayed perceptual breakthrough relative to stable-intensity conditions. However, when interocular conflict was imperceptible—thus minimizing executive involvement (Experiment 2a)—decreases in intensity enhanced subsequent perceptual dominance. These results underscore the dynamic interplay between bottom–up sensory processing and top–down cognitive control in visual competition.

The observed bias toward increasing intensity aligns with established research on attentional capture, reflecting an adaptive mechanism by which sensory systems prioritize abrupt or strengthening signals signaling potentially critical environmental events (Franconeri & Simons, 2003; Öhman, Flykt, & Esteves, 2001; Yantis & Jonides, 1984). Predictive coding frameworks offer a compelling theoretical interpretation of this asymmetry (Friston, 2005; Rao & Ballard, 1999). This framework becomes particularly powerful when considering the nature of a suppressed stimulus. A stimulus outside of reported awareness is, by definition, in a state of low perceptual salience, analogous to being occluded or decayed. Therefore, a decrease in its physical contrast aligns perfectly with the brain's existing model of this weak or absent state, thus generating minimal prediction error. Conversely, a sudden increase in contrast represents a significant violation of this model. It signals that a supposedly weak stimulus is, in fact, strong and escalating, generating a substantial prediction error that is prioritized to update the perceptual model and drive the stimulus into awareness (Hohwy, Roepstorff, & Friston, 2008). Our findings thus highlight perceptual mechanisms capable of processing unreportable visual inputs in response to sensory deviations.

Neuroimaging evidence further supports our arguments regarding sensory and executive contributions. Early visual regions, notably primary visual cortex (V1), demonstrate robust sensitivity to unexpected sensory inputs even during suppression (Alink, Schwiedrzik, Kohler, Singer, & Muckli, 2010; Fang & He, 2005; Summerfield & Egner, 2009; Tong, 2003). Experiment 2, in which explicit task demands were minimized, reduced prefrontal executive control involvement, thereby emphasizing bottom–up sensory processing (Brascamp et al., 2015). In this scenario, any intensity change—regardless of direction—enhanced perceptual dominance, reflecting inherent sensitivity of early visual mechanisms to sensory novelty. Conversely, under reportable rivalry conditions, top–down modulation from prefrontal and attentional networks becomes more prominent, selectively amplifying responses to upward intensity changes while attenuating downward changes (Desimone & Duncan, 1995; Miller & Cohen, 2001). Thus, perceptual dominance in binocular rivalry emerges from a delicate balance between bottom–up sensory detection and top–down attentional prioritization, modulated by the perceptibility of stimulus changes.

Our findings have significant theoretical implications for existing binocular rivalry models. Traditional mutual inhibition and adaptation frameworks (Blake, 1989; Wilson, 2003) may not fully account for the dynamic asymmetric effects observed with intensity changes. Incorporating Bayesian inference principles (Knill & Pouget, 2004), hierarchical predictive coding (Hohwy et al., 2008), and attentional gain-control mechanisms (Li, Rankin, Rinzel, Carrasco, & Heeger, 2017) into rivalry models would provide a more comprehensive explanatory framework. Such refinements allow for dynamic comparisons that account for perceptual switches following unexpected intensity changes (Cao, Pastukhov, Aleshin, Mattia, & Braun, 2021). Future models should explicitly integrate both dynamic sensory-driven mechanisms and flexible cognitive modulation to better explain the complex perceptual phenomena observed in binocular rivalry.

Several limitations of the current study should be acknowledged, which also point to valuable avenues for future research. First, we acknowledge that the increments and decrements in contrast were not necessarily matched for perceptual salience; it is unknown whether an increase from 0.4 to 0.8 is perceptually more or less salient than a decrease to 0.2. However, the qualitative, bidirectional nature of the effect (facilitation vs. active delay) points to a genuine, direction-specific neural mechanism, even if the quantitative magnitude is influenced by the step-size difference. Future studies could precisely titrate the contrast changes for each participant to be perceptually matched, which would allow for a more direct quantification of the directional asymmetry. Second, we note that, although we took measures to promote stable fixation, eye movements were not recorded, and we cannot entirely rule out their potential contribution to the observed effects. Future work using eye-tracking could clarify this point.

Conclusions

In summary, this study reveals a fundamental asymmetry wherein increases in the intensity of unreportable or suppressed stimuli preferentially enhance perceptual dominance during binocular rivalry compared with decreases. The magnitude of this asymmetry is modulated by the conscious perceptibility of rivalry stimuli, highlighting the interactive roles of bottom–up sensory processing and top–down attentional control. Our findings underscore the importance of incorporating predictive coding principles and attentional modulation into existing theoretical frameworks, advancing our understanding of how the brain dynamically prioritizes sensory input to shape conscious visual experience.