Effects of interocular grouping demands on binocular rivalry

Abstract

Binocular rivalry (BR) is a visual phenomenon in which perception alternates between two non-fusible images presented to each eye. Transition periods between dominant and suppressed images are marked by mixed percepts, where participants report fragments of each image being dynamically perceived. Interestingly, BR remains robust even when typical images are subdivided and presented in complementary patches to each eye, a phenomenon termed interocular grouping (IOG). The objective of the present study was to determine if increasing grouping demand in the context of BR changes the perceptual experience of rivalry. In 48 subjects with normal vision, mean dominant and mixed percept durations were recorded for classic BR and IOG conditions with increasing grouping demands from two, four, and six patches. We found that, as grouping demands increased, the duration of mixed periods increased. Indeed, durations of dominant and mixed percepts, as well as percentage of time spent in dominant or mixed state, differed significantly across conditions. However, durations of global dominant percepts remained relatively stable and saturated at about 1.5 seconds, despite the exponential increase in possible mixed combinations. Evidence shows that this saturation followed a nonlinear trend. The data also indicate that grouping across the vertical meridian is slightly more stable than for the horizontal meridian. Finally, individual differences in speed of alternation identified during BR were maintained in all interocular grouping conditions. These results provide new information about binocular visual spatial integration and will be useful for future studies of the underlying neural substrates and models of binocular vision.

Introduction

Binocular rivalry (BR) refers to the alternating periods of dominance and suppression experienced when each eye is presented with sufficiently dissimilar monocular input (Blake, 1989; Blake & Logothetis, 2002; Tong, Meng & Blake, 2006; Sutoyo & Srinivasan, 2009). During visual dichoptic presentation (i.e., separately to each eye) of two non-fusible images, subjects report periods of alternation between perceptual awareness of each image. Periods of alternating percepts can be characterized as near complete visibility of one image, termed the dominant percept, whereas the second image is suppressed from conscious perception. When subjects experience a perceptual switch, the previously suppressed image becomes dominant in visibility. During BR, the stimuli presented to each eye remain unchanged, however, the perceptual awareness of the images cycle between instances of dominance and suppression. Furthermore, during these periods of transition, subjects experience instances of partial visibility of patches from both images, termed mixed percepts. In recent years, more importance has been attributed to measuring mixed percepts experimentally and determining their theoretical role in binocular vision models (Said & Heeger, 2013; Katyal, Engel, He, & He, 2016, Sheynin, Proulx, & Hess, 2019; Bock, Fesi, Baillet, & Mendola, 2019). 

Although early models of BR focused primarily on reciprocal inhibition between monocular neurons (i.e., V1 neurons that receive input from one eye only) (Blake, 1989), a wealth of subsequent evidence has supported a role of binocular neurons as well (e.g., Logothetis, Leopold, & Sheinberg, 1996; Wilson, 2003; Said & Heeger, 2013). Findings that highlight the likely functional role of binocular neurons reveal that modifications of BR stimuli to disrupt monocular information still yield robust alternations of perceptual dominance. In one such demonstration, the entire image is exchanged between the eyes at a fast rate, but normal perceptual rivalry remains intact (e.g., Logothetis et al., 1996; see also Christiansen, D'Antona, & Shevell, 2017). Likewise, another important type of rivalry, and the focus of the present work, is interocular grouping (IOG). In IOG, complementary patches from two intact images are presented to each eye, requiring combinatorial processing between the visual inputs from each eye to achieve a unified percept of the global image (Kovács, Papathomas, Yang, & Fehér, 1996). Early observations in visual psychophysics revealed that complementary concentric circles and parallel lines (Figure 1A) yielded not only periods of alternations of so called “eye-based” monocular images but also unified percepts requiring grouping between the monocular inputs from the two eyes (Diaz-Caneja, 1928; translated by Alais, O'Shea, Mesana-Alais, & Wilson, 2000). In this way, evidence from IOG experiments provide compelling evidence that binocular neurons contribute to perceptual dominance and suppression in the context of BR. 

Figure 1.

 

Example of interocular grouping stimuli adapted from literature review. (A) Reproduction of Diaz-Caneja stimuli (1928), where stimulus flicker altered interocular grouping during BR (Knapen et al., 2007). (B) Reproduction of visual stimuli presented by Sutoyo and Srinivasan (2009). (C) Reproduction of visual stimuli presented by Golubitsky et al. (2019) showing four quadrant interocular grouping (see Discussion section).

Different proposals have been made to explain IOG. On the one hand, higher level stimulus pattern integration argues against monocular eye-dominance accounts of alternation during BR, as outlined in the previous paragraph (Kovács et al., 1996). Conversely, some proponents of lower level eye-of-origin theory claim that simultaneous eye-based dominance of local patches of the stimulus could drive alternating percepts during IOG (as might be implemented by horizontal connections between monocular neurons in V1) (Lee & Blake, 2004). In their view, coordinated dominance between difference regions of the retinotopic representation (local zones that correspond to local patches of the stimulus) as opposed to dominance based on the representation of the entire stimulus could explain IOG. However, recent functional magnetic resonance imaging (fMRI) findings support the idea that IOG rivalry primarily occurs between binocular neurons with larger receptive fields in higher level areas (Buckthought, Kirsch, Fesi, & Mendola, 2021). In sum, the observation that individuals experience coherent percepts when viewing intermingled patchwork of rivalrous images highlights the importance of spatial integration mechanisms for the experience of BR. 

In the context of BR with suprathreshold stimuli, mixed percepts rarely appear as a fusion or superimposition of the right and left eye stimuli (a plaid in the case of gratings). Instead, patches of partial visibility of each image are commonly reported (Kovács et al., 1996) and are also referred to as piecemeal rivalry (Alais & Melcher, 2007; Skerswetat, Bex, & Baron-Cohen, 2022). In the case of IOG, one would expect that periods of mixed percepts would be substantially increased if monocular eye-based rivalry was the primary mechanism. Alternatively, if binocular neurons are the primary substrate, then Gestalt principles of good continuation and global consistency may result in mixed percepts that minimally differ from classic BR. Accordingly, for both BR and IOG, characterizing mixed percepts is interesting, not only because they mark the intermittent transition period between cycles of dominance but also because they may represent a distinct mode of binocular integration that is crucial to the combination of information from both eyes during IOG. In fact, previous reports indicate that levels of inhibition in some cortical areas may increase during mixed percepts (e.g., Katyal at al., 2016). If IOG provides a more challenging environment for binocular integration to occur, we aimed to ask if mixed percepts would be selectively affected by increasing IOG demands. 

In recent years, tristability models of BR (i.e., alternations among three percepts) have emerged (Riesen, Norcia, & Gardner, 2019; Qiu, Caldwell, You, & Mendola, 2020). This view proposes alternations between each of the monocular inputs, as well as some type of mixed percept arising from the combination of the dichoptic stimuli. The model of tristability is relevant because it highlights mixed percepts as a distinct perceptual state of interest with its own relative stability. This emergent viewpoint follows in part from neurophysiological evidence linking intermittent periods of mixed percepts to a distinct neural signature of intermodulation frequencies when the images are frequency tagged during neuroimaging studies (Katyal et al., 2016; Bock et al., 2019). This outcome emphasizes the role of binocular neurons marked by the integration of monocular inputs, although these studies looked at intermodulation frequencies during BR only, and the same is not known for IOG. We aimed to explore the notion of tristability in classic BR and IOG with the inclusion of mixed percepts as a stable perceptual state. An indication of tristability in our study would find relatively proportional probability of transitions between dominant-to-dominant percepts, as well as mixed-to-dominant states. 

Alternation rates have long been important in the characterization of BR and other bistable percepts. This is defined as the number of perceptual alternations between each eye's stimuli over a defined time (Brascamp, Klink, & Levelt, 2015). Interestingly, there are many recent findings highlighting individual differences in switch rates observed during rivalry. The idea that switch rate is partially heritable is supported by twin studies (Miller et al., 2010; Shannon, Patrick, Jiang, Bernat, & He, 2011) and studies that found neural differences between fast and slow switchers (Fesi & Mendola, 2015; Bock et al., 2019). In addition, there is evidence that rivalry dynamics change with age, are affected by different neurological conditions or other previous experience (Scocchia, Valsecchi, & Triesch, 2014; Robertson, Ratai, & Kanwisher, 2016; Ye, Zhu, Zhou, He, & Wang, 2019). Finding a strong correlation between BR and IOG would indicate a similar underlying neural mechanism. 

In the current study, we compared the effects of increasing the number of patches distributed to each eye by dividing our classic BR image with orthogonal gratings into five different IOG conditions with two-, four-, and six-patch conditions (Figure 2). Grouping across the horizontal meridian was compared with grouping across the vertical meridian. Reports of mixed percepts as well as coherent dominance/suppression were obtained throughout all conditions. Whereas previous studies have used similar but isolated stimuli (Figures 1A and 1B), our design sought to systematically alter the location and number of image divisions while importantly maintaining the same visual percept for participants between classic BR and our five IOG conditions. Finally, in this study, we used flickering stimuli primarily to accommodate the separate magnetoencephalography experiments conducted with a subset of subjects (not reported here). We already know that dichoptic flicker alone does not cause rivalry. However, it is interesting that flickering stimuli have been shown to increase the duration of dominant percepts during interocular grouping with the use of flicker rates from 10 to 24 Hz (Knapen, Paffen, Kanai, & van Ee, 2007). For that reason, we also compared two rates of dichoptic flicker ranging between 5 and 12 Hz, to test for any large facilitatory effect on grouping when the higher flicker rate was used. 

Figure 2.

 

Experimental design and BR stimuli with increasing interocular grouping demands. (A) Experimental design to capture behavioral responses for alternations in percepts during BR and interocular grouping. Subjects were asked to report with a two-button press-and-hold response when viewing dominant percepts. Mixed percepts were inferred from periods of no responses. Illustration is shown over the time course of the visual presentation of rivalrous images. (B) BR with dichoptic presentation of stimuli. (C) Interocular grouping with two patches divided along the vertical meridian of each classic BR image. (D) Interocular grouping with two patches divided along the horizontal meridian of each classic BR image. (E) Interocular grouping with four patches divided along the vertical and horizontal meridian of each classic BR image. (F) Interocular grouping with six patches divided along the vertical meridian of each classic BR image. (G) Interocular grouping with six patches divided along the horizontal meridian of each classic BR image.

Methods

Participants

Participants recruited for inclusion reported normal or corrected normal vision and no known visual disorders. Contact lenses were allowed for visual acuity correction, but not glasses because of the prism glasses required for testing. We screened participants for visual impairments in acuity and stereo vision based on two assessments: the Logarithmic Visual Acuity chart (Early Treatment Diabetic Retinopathy Study [ETDRS] 2000 series chart; Precision Vision, Woodstock, IL), and the Titmus Stereoacuity Test (Stereo Optical, Chicago, IL). Participants completed these assessments in a well-lit room at the distance recommended by the testing manufacturer. Each eye was tested independently in visual acuity due to the dichoptic nature of BR. For each eye, visual acuity cut-offs for inclusion were 20/40 and no more than two lines of difference between eyes to avoid an underlying eye dominance. The median value for both right and left eyes was 20/20. For the stereo test, the inclusion criterion was set at 7/9, which corresponds to an angle of stereopsis of 60 seconds at 16 inches. The median results of the stereo vision test for the subject pool were 9/9. Out of the 61 participants recruited, nine did not meet the vision test criteria for inclusion and were excluded. The remaining 52 participants performed the experiment. Four participants were later excluded from the analysis due to testing values that were invalid for reasons beyond the scope of the experiment and were deemed not to represent the intended perceptual rivalry of the current study. For the remaining 48 participants, rare runs where they reported 100% mixed perception, periods of dominance below ∼100 ms (7 frames on the 60-Hz display), and self-reports of loss of fusion due to glasses fogging were removed. The final subject pool was N = 48, including two authors (EM and JM); 33 were females, and the mean age was 25.1 years with a standard deviation of 5.1 years. 

Experimental design

BR was achieved with the use of a black opaque divider placed between the two eyes and prism lenses 12 diopters in strength to overlay the images seen by each eye. When fusion was achieved by the participant on the fixation screen, the participant was instructed to self-initiate each testing block. All participants were guided through an unrecorded practice session (45 seconds for BR and IOG conditions) before the experimental conditions began. 

Throughout the experiment, participants were instructed to report alternations in dominant percepts (red or green) using a two-button press-and-hold design. Specifically, when the majority (80% or more) of the image was perceived to be red, participants were instructed to press the right-sided button at the onset of the dominant red perception and to continuously hold the key down throughout the stability of this percept. Alternatively, they were instructed to press and hold the left key for green percepts. When neither red, nor green was perceived as perceptually dominant, participants were instructed to withhold a response, indicating the experience of a mixed percept (Figure 2A). These instructions were identical for BR and IOG conditions. 

Stimuli

Images were shown using Psychtoolbox-3 (Brainard, 1997; Kleiner, Brainard, & Pelli, 2007) and MATLAB 2016a (MathWorks, Natick, MA), as illustrated in Figures 2B to 2G, with color codes in RGBA values for red [0.5 0 0 0.2] and green [0 0.4 0 0.2]. Pilot testing was conducted with five participants to achieve approximately balanced ratios for the proportion of time spent viewing red and green percepts. A fixation mark designed as a combination between a bullseye and crosshair was chosen to promote the stability (Thaler, Schütz, Goodale, & Gegenfurtner, 2013) required for interocular grouping. 

The stimuli shown had a visual angle of 6.67 (6°41′0.83′), with a viewing distance of 47 cm and stimuli size of 5.5 × 5.5 cm. The interocular orientation difference between red and green gratings was set at 90° for all conditions. The experimental design used stimulus-based tagging for two dichoptically presented flicker regimes that were compatible with a 60-Hz display. Flicker frequencies followed a slow and a fast regime of 5 and 6.67 Hz and 10 and 12 Hz, respectively. The green gratings were tagged at frequencies of 6.67 or 12 Hz, with corresponding frequencies of 5 or 10 Hz for red. Throughout the experiment, the images were counterbalanced for the color and orientation of gratings shown to each eye to limit the effects of adaptation. 

The experiment was comprised of seven conditions, each shown twice in 4-minute runs divided into four blocks of 1 minute. The order of conditions shown was counterbalanced and randomized among participants. Specific conditions were BR, two-patch vertical IOG (2VM), two-patch horizontal IOG (2HM), four-patch IOG (4), six-patch vertical IOG (6VM), and six-patch horizontal IOG (6HM). 

Results

Effect of flicker frequencies

In the Methods section, we explained how all stimuli consisted of red or green gratings, and, because we intended to use this experimental design for brain imaging, we also used stimulus-based tagging for two dichoptically presented flicker regimes (e.g., Bock, Fesi, Da Silva Castenheira, Baillet, & Mendola, 2023). The slow flicker regime used frequencies at 6.67 Hz and 5 Hz, whereas the fast flicker regime used 12 Hz and 10 Hz. Hence, we plotted the mean durations of dominant (red or green) percepts for BR and all IOG conditions combined and separately for slow and fast flicker frequencies (Figure 3). Dominance durations are computed as the average duration of red and green percepts. Overall, results showed that rivalry was slightly slower (longer percept durations and decreased alternation rate) for the slow flicker condition, as indicated by a Welch t-test that revealed a significant different difference between flicker speed for both BR (slow, 1.96 s; fast, 1.79 s; t = 2.32; df = 368; p = 0.02) and IOG (slow, 1.52 s; fast, 1.46 s; t = 2.35; df = 1828.6; p = 0.02). Next, mean durations of the mixed percepts were considered for BR and IOG. In this case, a Welch two-sample t-test was not significant for BR (slow, 1.08 s; fast, 1.06 s; t = 0.21; df = 364.04; p = 0.8) but was significant for IOG (slow, 2.39 s; fast, 2.21 s; t = 2.8; df = 1816.6; p = 0.005). Next, we examined alternation rate, which was computed as the number of perceptual alternations per second. Mean alternation rates were slightly higher for both BR and IOG with the faster flicker rate. Welch two-sample t-test was significant for BR (slow, 0.40 Hz; fast, 0.43 Hz; t = −2.57; df = 366.09; p = 0.01) and for IOG (slow, 0.32 Hz; fast, 0.34 Hz; t = −3.40; df = 1828.9; p = 0.0007). As observed across the results in Figures 3A to 3C, flicker frequency had a greater impact during BR, where the slower regime produced longer dominant percept durations and slower alternation rates; however, during IOG, increasing grouping demands led to a reduction in differences between the mean results, with the largest differences obtained during IOG2 with trends that mimicked BR. These results show that we obtained stable rivalry for both rates of flicker with only minor differences between them (considered further in the Discussion section). In all of the following analyses, with the exceptions of Tables 1 to 4, slow and fast flicker were combined in Figures 4 to 7. 

Figure 3.

 

Comparison of flicker frequencies. (A–C) Mean duration values for dominant and mixed percepts and alternation rate for BR and IOG. (D–F) Mean duration values for dominant and mixed percepts and alternation rate for increasing grouping demands during IOG. Dominant percepts are plotted as mean values for red and green responses. Black lines are the median values, and the dots represent participant data.

Table 1.

 

Multiple linear regression analysis between binocular rivalry and interocular grouping for mean dominant percept durations. Mean dominant percept duration was computed as the average time each dominant percept (red or green) was held by participants.

 

Figure 4.

 

Mean duration and proportion of viewing time for BR and interocular grouping. (A) Mean durations of percepts plotted for BR and IOG conditions. Duration of percepts is plotted as the mean duration for red and green percepts. (B) Mean proportion of viewing time for red, green, and mixed percepts across BR and IOG conditions. (C) Alternation rates plotted for each condition. Error bars plotted in all three graphs are 95% confidence intervals.

Effect of increasing interocular grouping demands

Percept durations

The mean duration of each sequential percept corresponds to perceptual stability in conscious perception of alternating percepts. Here, we computed mean percept durations for red and green percepts (labeled dominant), as well as mixed percepts (Figure 4A). Dominant percepts are most stable (longer duration) for classic BR (1.87, SD = 0.69), with a mean duration of nearly 2 seconds, and mixed percepts lasting on average just over 1 second (BR = 1.07, SD = 0.77). In contrast, for all the IOG conditions, the mean duration of dominant percepts (∼1.5 s) was shorter than BR (2VM = 1.59, SD = 0.58; 2HM = 1.47, SD = 0.53; 4 = 1.47, SD = 0.56; 6VM = 1.45, SD = 0.58; 6HM = 1.45, SD = 0.61), and less than the mean duration of mixed percepts for each IOG condition (2VM = 1.95, SD = 1.17; 2HM = 2.02, SD = 1.13; 4 = 2.61, SD = 1.48; 6VM = 2.35, SD = 1.44; 6HM = 2.61, SD = 1.53). However, it is notable that the dominant percepts remained very similar in duration even as grouping demands increased dramatically. In contrast, mixed percepts increased modestly as grouping demands increased. In addition, we see that mixed percepts are slightly longer for 2HM and 6HM (Figure 4A) than the image meridian division of 2VM and 6VM. In sum, between classic BR and interocular grouping conditions, there was an inversion of the duration for the most predominant type of percept (i.e., dominant or mixed) as seen in Figure 4A, with the experience of BR favoring longer dominant percepts, whereas IOG resulted in longer durations of mixed percepts. 

We performed a multiple linear regression analysis to examine whether there is an interaction between grouping demands and percept duration. There was a significant interaction in regression analysis with increasing grouping demands during IOG (p < 0.001; estimate = −0.12; confidence interval [CI], −0.15 to −0.09) and meridian division orientation (p < 0.001; estimate = −0.27; CI, −0.38 to −0.17), as reported in Table 2, favoring grouping in images divided along the vertical meridian. Therefore, grouping was affected by both the increase in number of patches and the orientation of their division. To determine whether mixed percepts differed between BR and IOG, we performed a multiple linear regression analysis that showed a main effect between the two conditions in Table 3 (p < 0.001; estimate = −1.35; CI, −1.51 to −1.19). Participants spent significantly longer periods viewing mixed percepts during IOG conditions. 

Table 2.

 

Multiple linear regression analysis for increasing interocular grouping demands across the number and orientation of image division. Mean dominant percept duration was computed as the average time each dominant percept (red or green) was held by participants. Increasing interocular grouping demands was defined as moving from IOG2, IOG4, to IOG6. Orientation of image division was either along the vertical or horizontal meridian.

 

Table 3.

 

Multiple linear regression analysis between binocular rivalry and interocular grouping for mean mixed percept durations and alternation rate. Mean mixed percept duration was computed as the average time each mixed percept was held by participants.

 

Percept predominance

The next step was computing the percentage of time each of the three possible percepts was reported, combining the two runs (slow and fast flickering rates) for each condition. The proportion of viewing time was calculated by taking the total duration of each condition measured in frames and dividing it by the total duration that participants indicated red, green, or mixed percepts. The predominance results provide a global view of the rivalry experience across the whole of the experiment and are plotted separately for red grating dominant, green grating dominant, or a mixed percept, for classic BR and each grouping condition (Figure 4B). The pattern of results is similar to the percept durations seen in Figure 4A. For classic BR, a dominant percept was perceived approximately 70% of the time. During BR, participants experienced the three percepts in the following viewing proportions: BR red = 0.40 (SD = 0.12); BR green = 0.34 (SD = 0.11); and BR mixed = 0.27 (SD = 0.19). During IOG conditions, the viewing proportions of the red grating (2VM = 0.29, SD = 0.11; 2HM = 0.27, SD = 0.11; 4 = 0.25, SD = 0.12; 6VM = 0.27, SD = 0.12; 6HM = 0.25, SD = 0.12) were seen slightly more often that green (2VM = 0.24, SD = 0.09; 2HM = 0.22, SD = 0.09; 4 = 0.19, SD = 0.09; 6VM = 0.19, SD = 0.09; 6HM = 0.18, SD = 0.09) which may be due to a small residual advantage in visibility, or even a bias at more cognitive levels. For the remaining viewing proportions, the proportions of mixed percepts were BR = 0.27 (SD = 0.19); 2VM = 0.47 (SD = 0.18); 2HM = 0.50 (SD = 0.18); 4 = 0.56 (SD = 0.19); 6VM = 0.54 (SD = 0.19); and 6HM = 0.56 (SD = 0.20), seen as a piecemeal percept, which was in line with previous studies that recorded mixed percepts. Sheynin et al. (2019) reported a proportion of mixed percepts for a subset of their participants with normal vision at approximately 42%. Given that our stimuli differed in color, spatial frequency, and size, our results are reasonably comparable with their findings. When IOG is next considered, it is evident that mixed percepts are perceived for a significantly longer period (Figure 4B), in the range of approximately 45% to 55%. Dominant percepts were perceived significantly less than for classic BR. In comparison, Kovács et al. (1996) originally reported 60% dominant and 37% mixed viewing, with the remaining time spent viewing fused percepts, although direct comparison is somewhat challenging due to differences in the stimuli used. As our grouping demands increased, there was a rather modest increase in mixed percepts from IOG2 to IOG6 (Figure 4B). However, it is notable that we see a plateau that is better modeled by a nonlinear, second-order polynomial function, rather than a linear function in the percentage of viewing time for mixed percepts, even as the number of patches increases, and the number of possible mixed percepts increases exponentially (further considered in the Discussion section). Finally, we also found that rivalry was slightly more stable when grouping was required across the vertical meridian compared with the horizontal meridian, defined by orientation of the central image division. 

Alternation rate

Alternation rate was measured in hertz as the number of alternations in dominant percepts (red and green) per second. Higher alternation rates are indicative of faster switching between red and green dominant percepts experienced by the participant and are often cited as “fast switcher” in rivalry literature (e.g., Fesi & Mendola, 2015; Bock et al., 2019). The alternation rate was calculated by dividing the total duration of time spent viewing each condition by the number of red and green button responses (indicated switch of percept). We found that classic BR created the fastest rivalry experience (BR = 0.42, SD = 0.13), and IOG always results in a slower mean rate of dominance alternations (2VM = 0.35, SD = 0.12; 2HM = 0.35, SD = 0.12; 4 = 0.30, SD = 0.11; 6VM = 0.33, SD = 0.12; 6HM = 0.31, SD = 0.11) as shown in Figure 4C. In other words, rivalry was slowed down under conditions of interocular grouping because subjects spent more time in (mixed percept) transitions periods. 

Furthermore, the multiple linear regression analysis shown in Table 4 was performed to investigate the relationship between increasing grouping demands and alternation rate. This analysis resulted in a significant interaction between alternation rate and the number of patches in IOG (p < 0.001, estimate = 0.03 Hz; C, −0.03 to −0.02). We observed a decrease in alternation rate as the grouping demands increased. 

Table 4.

 

Multiple linear regression analysis between binocular rivalry and interocular grouping for alternation rate. Alternation rate was defined as the number of switches between dominant percepts per second. It was computed by dividing the total number of alternations between dominant percepts by the duration of the experiment.

 

Individual differences in switch rates and percept durations

We performed correlation analyses between the BR and IOG conditions to observe the relationship between conditions for percept durations (dominant and mixed) and alternation rate. The correlations for alternation rate between conditions was robust and high in the positive direction, as shown by the R values in Figure 5C. This indicates that fast switchers for BR were also faster switchers during IOG. Although rate is traditionally used as a primary measure to gauge rivalry dynamics, we also investigated the correlations between both mean durations for dominant and mixed percepts (i.e., the duration percepts were held). The correlations between conditions for dominant percept durations (Figure 5A) and mixed percept durations (Figure 5B) were also found to be strong and positive as shown by the R values shown in Figure 5. All correlations were found to be significant after false discovery rate (FDR) correction when using the p.adjust function in R (R Foundation for Statistical Computing, Vienna, Austria). Within IOG, a strong positive relationship was also found among the conditions. Simply put, relationships were strongest for images that were closer in their number of divisions (i.e., IOG2 and IOG4, or IOG4 and IOG6), whereas conditions with greater dissimilarity (i.e., IOG2 and IOG6, or BR to IOG4/IOG6) had comparatively decreased R values. Taken together, the alternation rate, mean duration for dominant, and mean duration for mixed percepts provide an overall view of the rivalry dynamics experienced by participants and highlight the strong positive correlations found (Figure 5) among the conditions. 

Figure 5.

 

Correlation matrix with R values plotted of mean values for participants depicting relationship between BR and interocular grouping. (A) Mean duration (seconds) of dominant (red and green) percepts. (B) Mean duration (seconds) of mixed percepts. (C) Alternation rate (Hz). All correlations were statistically significant after FDR correction.

The results in Figure 5 were calculated for 95% CIs, with the following results. For dominant percept durations, BR-IOG2 = 0.73 to 0.90, BR-IOG4 = 0.57 to 0.79, BR-IOG6 = 0.61 to 0.83, IOG2-IOG4 = 0.89 to 0.95, IOG2-IOG6 = 0.83 to 0.94, and IOG4-IOG6 = 0.90 to 0.96 (Figure 5A). For mixed percept durations, BR-IOG2 = 0.59 to 0.79, BR-IOG4 = 0.26 to 0.57, BR-IOG6 = 0.47 to 0.72, IOG2-IOG4 = 0.68 to 0.88, IOG2-IOG6 = 0.73 to 0.91, and IOG4-IOG6 = 0.73 to 0.89 (Figure 5B). For alternation rate, BR-IOG2 = 0.83 to 0.92, BR-IOG4 = 0.49 to 0.79, BR-IOG6 = 0.56 to 0.84, IOG2-IOG4 = 0.64 to 0.89, IOG2-IOG6 = 0.70 to 0.90, and IOG4-IOG6 = 0.82 to 0.94 (Figure 5C). 

In a further analysis, illustrated in Figure 6, we sought to separately evaluate whether dominant or mixed percepts were better correlated with alternation rates across BR and IOG. During BR, we found a strong negative correlation between both the mean dominant percept duration and alternation rate (R = −0.55, t = −4.5204, df = 46, p = 4.307e-05) and mixed percept duration and alternation rate (R = −0.48, t = −3.6803, df = 46, p = 0.0006097), plotted in Figure 6A. However, this was not the case during IOG (Figure 6B), where only the duration of mixed percepts was found to show a significant negative correlation with alternation rate (R = −0.84, t = −10.426, df = 46, p = 1.062e-13), whereas there was a nonsignificant relationship with the duration of dominant percepts and alternation rate (R = −0.21, t = −1.4628, df = 46, p = 0.1503). Fast switchers in both BR and IOG conditions tended to have shorter mixed percepts, but for dominant percepts fast switchers were significantly quicker only during BR, and not IOG, whereas slow switchers tended to have longer durations of mixed percepts. 

Figure 6.

 

Scatterplot depicting correlations between the alternation rate and mean duration of mixed percepts, with linear lines of best fit. (A) BR alternation rate and mean duration of dominant and mixed percepts. (B) Interocular grouping alternation rate and mean duration of dominant and mixed percepts.

Transition probabilities between perceptual states as an indication of tristability

As previously proposed by Riesen et al. (2019) and Qiu et al. (2020), models of tristability during classic BR and IOG were tested. Our analysis used transition probabilities between perceptual states during rivalry. Transition probabilities between perceptual states reported by participants were computed by comparing the current percept reported and their prior state of perception. With the inclusion of mixed percepts as a behavioral response, participants were able to report three different percepts: red, green, and mixed (Figure 7). When viewing a percept, there was the possibility of transitioning to either of the two different states, such that a participant viewing mixed could report the transition toward either a red or a green percept. For this reason, the arrowed lines leaving a percept in Figure 7 are equal to the value of 1 (i.e., the arrows leaving a green percept toward both a red and mixed percept) and point in the direction of the percepts reported. For the purposes of this computation, stable percept periods were defined as those lasting more than 100 ms in duration to avoid mislabeling the transitions between dominant to dominant as including a mixed period when participants were shifting between response keys. 

Figure 7.

 

Transition probabilities between perceptual states during rivalry. (A) Results for BR. (B) Combined results for all conditions of interocular grouping. Lines between perceptual states are drawn to scale of the mean probability of values of transition. Arrows point toward percepts reported from the previous state.

For both BR and IOG, there are similar transition probabilities when moving from a mixed to dominant percept, with a slight preference toward green. These values are taken to be 0.46 (SE = 0.008) and 0.43 (SE = 0.003) from mixed to red and 0.54 (SE = 0.008) and 0.57 (SE = 0.003) from mixed to green for BR and IOG, respectively. When viewing a red percept, participants had a greater tendency to transition to a mixed percept, with probabilities of 0.66 (SE = 0.017) for BR and 0.86 (SE = 0.005) IOG, whereas it was less often reported to move between red to green with values of 0.34 (SE = 0.017) for BR and 0.14 (SE = 0.005) for IOG. When experiencing green, a similar trend was observed with transition probabilities of BR and IOG going to mixed of 0.68 (SE = 0.017) and 0.87 (SE = 0.005), with a decline of values observed moving to red at 0.32 (SE = 0.017) and 0.13 (SE = 0.005) for BR and IOG, respectively. 

Importantly, between BR and IOG, there was a difference in the transition probabilities of moving between dominant-to-dominant percepts. During BR, participants reported moving between red and green roughly a third of the time, whereas this value was greatly reduced during IOG. In addition, the results between IOG conditions did not vary greatly for dominant-to-dominant transition probabilities from red to green (IOG2 = 0.16, IOG4 = 0.13, IOG6 = 0.13) and from green to red (IOG2 = 0.15, IOG4 = 0.12, IOG6 = 0.12). This result allows us to understand the tristable nature of the rivalry dynamics experienced by participants during BR. In contrast, there was far greater likelihood of participants reporting mixed percepts after transitioning from either red or green percepts during IOG (Figure 7B). 

We found no evidence that flicker rate impacted the nature of this tristability during BR. During binocular rivalry, slow and fast flicker regimes yielded similar transition probabilities. For mixed to red (slow, 0.47; fast, 0.45), mixed to green (slow, 0.53; fast, 0.55), red to mixed (slow, 0.66; fast, 0.66), green to mix (slow, 0.69; fast, 0.67), red to green (slow, 0.34; fast, 0.34), and green to red (slow, 0.31; fast, 0.33). 

Finally, a related phenomenon known as switchbacks was also calculated. Switchbacks represent the experience of going from a dominant to mixed and back to the previously reported dominant state (i.e., red, mix, red perceptual reports). The probability of switchbacks from red–mix–red were as follows: BR = 0.07, IOG2 = 0.11, IOG4 = 0.13, and IOG6 = 0.12. Similarly, for green–mix–green, BR = 0.11, IOG2 = 0.21, IOG4 = 0.27, and IOG6 = 0.27. Thus, a trend was observed for higher switchback probability during IOG when compared with BR. This result further supports the evidence for a tristable regime observed during BR with its lower probability of switchbacks. 

Discussion

Our study on visual rivalry sought to explore the behavioral outcomes (i.e., mean dominant and mixed durations, alternation rate) when the number of complementary grouped patches during interocular grouping was systematically increased, from classic BR to two, four, and six image divisions. Classic BR resulted in the longest mean dominant percept durations, which were roughly twice as long as the mean for mixed percepts. This trend was reversed for the grouping conditions (Figure 4A), with longer mixed percept lengths relative to the dominant percepts. In fact, the mean dominant durations for IOG showed remarkable stability as the number of complementary grouped patches increased; apart from 2VM (1.59 s), the grouping conditions of 2HM, 4, 6VM, and 6HM ranged from only 1.45 to 1.47 seconds. This is an important result, because it highlights the stability of dominant percepts as the demands of interocular grouping increase. In contrast to dominant percept durations, as grouping demands increased, the mean mixed percept duration increased from the two- to four-patch grouping conditions. A nonlinear trend was observed when moving from the four- to six-patch conditions, better modeled by a nonlinear, second-order polynomial function than a linear equation. We thus found that alternation rates decreased with increased grouping demands and showed that individual differences in switch rates for classic BR are maintained with IOG. Finally, the preponderance of mixed percepts in IOG resulted in a departure from the classic BR pattern that better resembles tristability, with a dominant (red/green) percept equally likely to transition to mixed or dominant (green/red). In IOG, dominant percepts usually transition to mixed before dominance is again established. 

Impact of flicker rate in the experience of rivalry

The design of our study sought to test whether altering the dichoptic flicker regime between slow (5 and 6.67Hz) and fast (10 and 12 Hz) frequencies affected the experience of BR and interocular grouping. In both BR and IOG, we observed similar trends; modest increases in both dominant and mixed percept duration for slow flicker compared with fast flicker (Figure 3). These results slightly extend the findings by Knapen et al. (2007) to lower frequency regimes; they previously showed that that flickering images (>10 Hz) increased interocular grouping compared with static images. We were able to determine that slower rates of flicker did not adversely disrupt IOG and may even stabilize it. This is of practical importance for neuroimaging studies of visual rivalry that rely on frequency tagging below 10 Hz (e.g., Tononi, Srinivasan, Russell, & Edelman, 1998; Srinivasan, Russell, Edelman, & Tononi, 1999; Cosmelli et al., 2004; Kamphuisen, Bauer, & van Ee, 2008; Sutoyo & Srinivasan, 2009; Jamison, Roy, He, Engel, & He, 2015; Katyal et al., 2016; Roy, Jamison, He, Engel, & He, 2017; Bock et al., 2019). The difference noted between our slow (5 and 6.67 Hz) and fast (10 and 12 Hz) was significant but small in magnitude for the measures of mean durations and alternation rate. Therefore, a wide range of frequencies appears to provide adequately long durations of perceptual stability between alternations, which can be advantageous in steady-state visually evoked potential neuroimaging studies. 

Importance of mixed percepts

Although a few previous studies of IOG exist, none of them measured mixed percepts as systematically as the current study (Knapen et al., 2007; Sutoyo & Srinivasan, 2009; Golubitsky, Zhao, Wang, & Lu, 2019). Our study was designed to explicitly measure the experience of mixed percepts, without adding additional high cognitive demands to traditional two-button response studies. We were able to further characterize mixed periods in their frequency, duration, proportion of viewing time, correlation to switch rates, and transition probabilities. The periods where (ungrouped) portions from images shown independently to each eye are visible represent ambiguity and competition within the visual system prior to the resolution of coherent visual perception. Although the perceptual experience of BR was well mimicked during IOG, fundamentally the two visual illusions differ. During BR, alternations took place between monocularly driven stimuli, and the perception of mixed percepts represented periods where both images demonstrated partial visibility. In contrast to this, the experience of IOG involves each eye being shown a version of a so-called “static” mixed percept, and it is up to the neural mechanisms of the visual system to elicit the seamless experience of dominance and suppression as is the case during BR. It is thus important to mention that, although we observed an inversion in the durations of dominant and mixed percepts between BR and IOG, dominant durations during IOG were significantly longer than mixed percepts during BR (non-overlapping error bars in Figure 4A). These periods of dominance during IOG may represent durations where combinatorial processing between binocular driven neurons is occurring. We note in passing that the mixed percepts observed during IOG may entail competition between monocular images as well as various ungrouped binocular combinations, and thus the processing is not identical to that during BR. Comparing the tagged intermodulation frequency signals physiologically recorded for mixed percepts during BR with IOG would be informative in future studies (see Katyal et al., 2016; Bock et al., 2019). 

For IOG, it is interesting that the visual system “unscrambles” the mixed monocular inputs to mimic classical BR alternation experience so robustly. This highlights some of the Gestalt principles of similarity in color, continuity in oriented gratings, and pattern completion to drive perception toward a unified image. We observed a distinct dissociation between the majority of percepts perceived by participants between BR and IOG. For BR, dominant red/green percepts were reported in an approximate 2:1 ratio over mixed percept for their duration (Figure 4A) and approximately a third of viewing proportion (Figure 4B). This trend was reversed during IOG with an increasing duration of mixed percepts as the conditions increased in grouping demands, as well as an increase in the proportion of mixed percept viewing. To emphasize the remarkable stability of grouping mechanisms as demands increased, we observed a nonlinear trend in total mixed viewing proportion that plateaued below 0.6 (Figure 4B) even as the number of divisions was increased from two to six patches. This was previously modeled by Kovács et al. (1996) as the possible number of combinations that can be viewed based on the assumption that perception alternates randomly in each complementary patch (i.e., between red and green). The authors proposed that the probability of a uniform percept (p_u) can be modeled by the equation p_u = 2 × 1/2ⁿ, where n is the number of bistable patches. During the IOG4 conditions, where participants viewed a superimposed rivalrous quadrant with four bistable patches, p_u = 2 × 1/2⁴ results in 0.125 (12.5%). This result was pushed even further during IOG6, where p_u = 0.003125 (3.125%) and p_m (where m = proportion of mixed percepts) can be modeled as p_m = 1 – p_u = 0.96875 (96.875%). Clearly, this model would not explain our observation that participants never exceeded a group mean of 60% viewing proportions for mixed percepts as grouping demands linearly increased. 

Several previous studies on IOG limited stimuli to two or four patches (Knapen et al., 2007; Sutoyo & Srinivasan, 2009; Golubitsky et al., 2019). Our results correspond with these findings, despite differences in stimulus parameters that could promote global percepts. The seminal paper by Kovács et al. (1996) reported mixed proportions of 37% for BR and 50% for IOG, with limited evidence of binocular fusion. Subsequently, about of 40% viewing time spent in mixed periods has been cited during rivalry by Alais & Melcher (2007), and Sheynin et al., (2019). These results are in line with our findings from two-patch IOG, but we observed a further increase in the relative proportion of mixed percepts as the grouping demands increased. Moreover, we found that there was a strong negative correlation between alternation rate and duration of mixed percepts (Figure 6B) during IOG, but durations of dominant percepts were remarkably stable at approximately 1.5 seconds. This suggests to us that the factors and mechanisms determining a switch from dominance to suppression may differ from a switch from suppression to dominance (Bock et al., 2023). The former mechanism that determines the duration of dominance appears minimally affected by grouping demands and may rely more heavily on ventral stream visual areas (e.g., Sandberg, Bahrami, Lindelov, Overgaard, & Rees, 2011; Sandberg et al., 2014, Bock et al., 2023). The later mechanism, which establishes dominance and may be more reliant on dorsal stream visual areas according to some evidence, appears more vulnerable to grouping demands. 

Effect of increasing grouping demands

We defined difficulty in grouping as the number of divisions and complementary bistable patches used to elicit IOG during rivalry. Throughout the experiment, it was observed that participants viewed dominant percepts with greater ease during the two-patch IOG conditions, as opposed to the four- and six-patch IOG (Figures 3 and 4). Of interest, participants qualitatively described the experience of rivalry during higher order grouping conditions (e.g., IOG6) to build up in nature toward a dominant percept. They reported mixed percepts to be dynamically perceived, whereas the complementary bistable patches would flip toward a coherent percept and be stable in perception during the report of a dominant percept, similar to the experience of dominance during BR. Some subjects reported awareness of a gradual “wave” of dominance so that spreading of coherency occurred sequentially for adjacent patches. Interestingly, as we increased grouping demands during IOG, an observable difference was present between IOG2 to IOG4 and between IOG4 to IOG6. We found the experience of rivalry to be slowed in the four-patch condition when compared with IOG2, due to the reduced alternation rates, and increase in mixed percept durations. The same cannot be said of the transition from four to six bistable patches, which led us to hypothesize that grouping along both the vertical and horizontal meridian must be resolved for a global uniform percept. We can thus predict that the gatekeeping step in the alternation rate during IOG was the resolution of conflict across both the vertical and horizontal meridian. We could further hypothesize that increasing the number of complementary patches beyond what was tested may not have as great of an effect on alternation rate as the initial vertical and horizontal conflict resolution. 

Interocular grouping across the horizontal versus vertical meridian

We observed a clear difference in grouping across the vertical and horizontal meridians of the visual field that lead to a preference in what we defined as IOG stimuli divided across the vertical image meridian. Our findings can be related to a previous neuroimaging study (Large, Culham, Kuchinad, Aldcroft, & Vilis, 2008) that posits a two-stage model of spatial integration in the visual hierarchy of higher tier visual areas. Namely, visual representations for upper and lower visual representations are processed in the lateral occipital (LO) prior to integration across left and right visual hemifields. Similarly, Larsson and Heeger (2006) found a role for visual areas LO1 and LO2 in integration of shape information for the entire contralateral visual hemifield, including both upper and lower quadrants. Tootell, Mendola, Hadjikhani, Liu, and Dale (1998) hypothesized that information was processed in multiple stages, first in the contralateral visual field and then across the vertical meridian as the receptive fields increase in size and extend to ipsilateral visual field, in areas such as LO. With this model and the two-stage model proposed by Large et al. (2008), we could expect a difference between IOG grouping demands across the vertical or horizontal meridian and slower rivalry dynamics when increasing the number of bistable patches to IOG4. This was shown to be true for Golubitsky et al. (2019), who found differences in the perceptual reports between IOG2 and IOG4 patch conditions, although they had a limited subject pool of n = 3. For the condition labeled 2VM in our study, they found more reports of all red and green percepts during their 2-second IOG presentation experiment, whereas the 2HM stimuli produced more monocular percepts, meaning less evidence grouping during rivalry. The diagonal condition they used resembled our IOG4 condition and showed results similar to those for 2HM. This could be interpreted as grouping across the 2HM condition being a limiting factor in a two-step process. Our results are similar, with participants viewing more dominant and less mixed percepts during the IOG stimuli with vertical image meridian divisions. 

We note in passing additional anatomical and physiological evidence for how visual quadrants are combined that may impact vertical vs horizontal IOG grouping (Saint-Amour, Lepore, Lassonde, & Guillemot, 2004; Knyazeva, Fornari, Meuli, & Maeder, 2006). A particular area of interest is the lateral occipital visual cortex, previously found to be important for integration of information during interocular grouping with fMRI in the 2HM condition (Buckthought et al., 2021). The current behavioral results would predict a noticeable change in fMRI results between IOG2 and IOG4 due to the demands of integration across both the vertical and horizontal image meridian. When information from within and between hemifields is integrated, we could thus expect similar results as grouping demands increase from four to six patches. 

Effects of collinearity and end-stopping

Another important factor that may explain our results is the well-known faciliatory (excitatory) effects of colinear oriented stimuli in visual cortex (Polat & Sagi, 1993; Polat, 1999; Cass & Alais, 2006). As has been previously hypothesized, collinear facilitation may rely on different mechanisms involving both lateral connectivity within V1 and extra-striate feedback (Cass & Alais, 2006; Jachim, Gowen, & Warren, 2017). When we increase the number of complementary grouped patches, we would expect a decreased collinear interaction for the monocular images in the visual cortex that may weaken their representation at the level of V1. Increasing the number of divisions equally results in a smaller surface area for each of the complementary patches, which in turn increases the proportion of mid- to higher spatial frequencies presented. This outcome may shift the balance between lower and higher tier visual areas and between monocular and binocular competition, leading to greater mixed percepts and the small reduction in dominant percept durations observed. Another consideration is that, as we add patches, we increase the amount of end-stopping and potentially have a wider band of orientations (from patch edges) in the stimulus that will selectively stimulate different regions of the visual cortex. We presume that end-stopping as first presented by Hubel and Wiesel (1965) would have the opposite effect of collinearity. In primates, end-stopped neurons respond better to short as opposed to long contours (Pack, Livingstone, Duffy, & Born, 2003). The increased excitation of end-stopped neurons may have contributed to our nonlinear trends observed as participants moved to higher order grouping conditions. Finally, it is likewise possible that the differences observed in participants between lateral and feedback connections to V1 (Jachim et al., 2017) could also account for some of the individual differences in IOG observed during our study. 

Individual differences during classic BR and IOG

Individual differences in speed of alternation during BR have been well documented and are heritable (Miller et al., 2010; Scocchia et al., 2014). Moreover, physiological studies have also examined the distinction between fast and slow switchers (e.g., Fesi & Mendola, 2015; Bock et al., 2019) in order to better understand the sources of such trait-like behaviors. Our results show a strong positive correlation between the individual differences in alternation rates between BR and IOG. The relationship was strongest for stimuli that were most similar (i.e., BR to two patches) and decreased as the grouping demands increased (BR-IOG2, R = 0.89; BR-IOG4, R = 0.75; BR-IOG6, R = 0.68). This is consistent with the finding that mixed percept durations systematically increase at the group level as grouping demands increase. However, the fast switchers for both BR and IOG experienced relatively shorter mixed percept durations, whereas slower switchers had longer periods of mixed percepts. This is important because it highlights again the transition from suppression to dominance, this time as a distinction between fast and slow switchers. Fast switchers can transition to dominant percepts more easily, which may be due to lower levels of baseline inhibition in the visual system that provides stability of dominant and suppressed inhibited percepts. Further, neural mechanisms such as the balance between excitation and inhibition have already been linked to altered rivalry in the case of autistic traits where there is a hypothesized imbalance (Robertson, Kravitz, Freyberg, Baron-Cohen, & Baker, 2013; Robertson et al., 2016; Dunn & Jones, 2020). 

Evidence for tristability

BR has long been modeled as a bistable process, in line with other examples of bistable and multistable illusions (Blake & Logothetis, 2002; Rodríguez-Martínez & Castillo-Parra, 2018). However, the possibility of rivalry between more than two possible percepts has been recognized for some time (e.g., Said & Heeger 2013; Katyal et al., 2016; Riesen et al., 2019; Bock et al., 2019; Qiu et al., 2020). In fact, for specific classic BR conditions, binocular fusion or mixed percepts may exhibit a pattern of tristability, such as for gratings with a small difference in interocular orientation (Riesen et al., 2019), as well as for luminance contrast BR stimuli (Qiu et al., 2020). These models would assume that mixed percepts have relative stability in their durations and alternate in conjunction with the other percepts. In addition, a tristable illusion would entail approximately equal probabilities of transitioning to either of the two other percepts from the current perceptual state (Riesen et al., 2019). Our results for BR have an indication for tristability among red, green, and mixed percepts; however, this was not observed during interocular grouping. We show here specifically that switching through a period of mixed percepts was not mandatory for BR, whereas for IOG we found little evidence of transitions directly from one dominant percept to the other. 

This could point toward evidence for a difference in the mechanisms of alternation between the two visual illusions. Monocular competition may be more prevalent overall when experiencing BR, whereas binocular integration is crucial for the experience of IOG. Taken together, our results seem indicative of a similar network that might operate differently to accommodate grouping during BR. We already have some evidence that BR and IOG share a similar overarching “rivalry network” that differs mainly in higher order visual regions that have specificity for grouping demands. fMRI studies on IOG have found increased blood-oxygen-level-dependent (BOLD) signal in higher order visual areas such as the lateral occipital (LO) area and intraparietal regions (Buckthought et al., 2021). Moreover, electroencephalogram (EEG) and transcranial magnetic stimulation (TMS) studies have highlighted the role of the parietal cortex in perceptual reversals more generally (Zaretskaya, Thielscher, Logothetis, & Bartels, 2010; Pitts & Britz, 2011). In a proposed tristable model of BR (Figure 8), rivalry could be occurring not only between monocular driven neurons but also with binocular driven neurons. The resolution of this conflict in earlier regions (i.e., primary visual cortex) would require a higher order visual region that received information from both pools of neurons, resulting in the results observed during BR. In particular, interocular competition may act locally in V1 to establish patches of dominant populations of neurons that feed up to higher levels and interact with binocular feature-based competition in an area such as LO. Concurrently, LO feeds back signals that bias the lower levels toward global consistency, which in turn further promotes a winner at the higher level. This cycle may repeat and might take longer in the case of IOG (during the longer mixed percepts). Such a model might limit the piecemeal rivalry with a traditional stimulus and also allow global feature dominance to occur in IOG. 

Figure 8.

 

Visual model that accounts for tristable paradigm observed during BR. Schematic of competition occurring at higher level of the visual pathway, in a proposed region that received information from both monocular and binocular driven neurons. Competition in higher order visual areas results in perception of the dominant percept and feedback inhibition of suppressed percepts.

Taken together, our results illustrate the strength of the visual system in grouping dichoptic stimuli with similar color, orientation, and good continuation, even as the interocular grouping demands increased during rivalry. We found differences in grouping across the vertical and horizontal image meridian during IOG, and our results indicate a slight preference across participants for images divided along the vertical meridian. We further supported the notion of tristability during BR, which builds on models of rivalry to include rivalry between both monocular and binocularly driven neural mechanisms. Although we found mixed percepts to have an increased predominance during interocular grouping, the visual system was robust toward its preference for uniform percepts. Finally, of interest for the study of rivalry and visual illusions, we found a strong positive correlation between behavioral results during BR and IOG conditions, indicative of a similar neural mechanism at play dictating perception. 

Acknowledgments

The authors thank the helpful input of two anonymous reviewers. 

Commercial relationships: none. 

Corresponding author: Eric Mokri. 

Email: [e-mail]. 

Address: Department of Ophthalmology and Visual Sciences, McGill University, Montreal, QC, Canada.