Does V1 response suppression initiate binocular rivalry?

Summary During binocular rivalry (BR) only one eye’s view is perceived. Neural underpinnings of BR are debated. Recent studies suggest that primary visual cortex (V1) initiates BR. One trigger might be response suppression across most V1 neurons at the onset of BR. Here, we utilize a variant of BR called binocular rivalry flash suppression (BRFS) to test this hypothesis. BRFS is identical to BR, except stimuli are shown with a ∼1s delay. If V1 response suppression was required to initiate BR, it should occur during BRFS as well. To test this, we compared V1 spiking in two macaques observing BRFS. We found that BRFS resulted in response facilitation rather than response suppression across V1 neurons. However, BRFS still reduces responses in a subset of V1 neurons due to the adaptive effects of asynchronous stimulus presentation. We argue that this selective response suppression could serve as an alternate initiator of BR.

The role of primary visual cortex (V1) for binocular rivalry (BR) is unclear V1 population spiking is reduced at the onset of BR, providing a potential trigger However, this broad spiking suppression does not occur for a variant of BR The BR variant reduces subpopulation responses, a potential alternate trigger

INTRODUCTION
Seeing something different in each eye (dichoptic viewing) is a constant part of the visual experience. One reason that the two eyes' views are not identical is that the horizontal offset between the eyes causes a geometrical difference in perspective. 1 Each eye's view is also distinct because of differences in the obstruction by both the nose and the retinal blind spots. 2 Furthermore, the two peripheral visual fields can only be seen by either eye. 3 One of the main roles of the primary visual cortex (V1) is to resolve minor differences between each eye's view to produce fused (''cyclopean'') visual information to feedforward to the rest of the visual system. [4][5][6] However, larger differences between the two eyes' views cannot be reconciled. In this case, our visual system either resorts to double vision (diplopia) 7,8 or one eye's view dominates over the other (binocular rivalry). [9][10][11][12][13][14][15][16][17][18][19][20][21][22] Binocular rivalry is a fascinating phenomenon with wide-ranging implications, from neural mechanisms of binocular vision [23][24][25][26] to the neural correlates of consciousness. 11,17,[27][28][29][30][31] As a consequence, binocular rivalry has been studied extensively both on the psychophysical as well as on the neuronal level. 9,14,32-35 Despite the large body of work on binocular rivalry, there are still open questions about its neural basis.
Increasing proportions of neurons correlate with perceptual reports of binocular rivalry as the hierarchy of cortical visual areas is ascended. 14,31,[36][37][38][39][40][41][42][43][44][45][46] Notably, initial studies found that less than a quarter of neurons in V1 modulate with perception during rivalry. 42 In contrast, neuroimaging (fMRI) indicates that V1 modulates strongly with the perceptual state of the subject. 47,48 This finding holds even within subjects, and it is still unclear what causes the apparent discrepancy 49 Taken together, these findings suggest that neural responses associated with binocular rivalry occur across the entire visual system, 2,9,21 including -to some degree -V1. However, the precise role that V1 plays in this intricate network of rivalry-related activation is still unclear. It remains uncertain if V1 predominantly contributes to the detection of interocular conflict, the suppression of one eye's view, or the initiation of reversal between each eye's perspective.
One intriguing idea is that V1 may be needed to initiate rivalry. This concept has gained special importance as discrete stages of binocular rivalry -particularly onset rivalry -have been identified. 50-53 Furthermore, several computational models 54-60 and V1 neurophysiology results [61][62][63][64][65] (see next paragraph) have suggested that V1 may initiate rivalry through mechanisms of neural inhibition. Most importantly, fMRI has If V1 were to play a crucial role for the initiation of binocular rivalry, how could this be conducted at the cellular level? Several computational models propose that response suppression allows for instability in neural population responses, thus enabling the initiation of rivalry. 56, [73][74][75] A specific mechanism of V1 response suppression that could act as initiator of binocular rivalry is dichoptic cross-orientation suppression (dCOS). [61][62][63][64][65] dCOS refers to an unspecific reduction of V1 responses that occurs during conflicting binocular stimulation. 64 In other words, onset of binocular rivalry-inducing stimuli reduces V1 spiking responses across all types of V1 neurons, no matter their specific tuning properties or the perceptual state of the subject [61][62][63][64][65]76,77 (and see DeAngelis et al. 78 ). For ease of communication, we will refer to this phenomenon as response suppression (to dichoptic stimuli).
One reason that response suppression to dichoptic stimulation is interesting is that the onset of response suppression roughly coincides with the perceptual onset of binocular rivalry. 52,53,61,63 This observation can be seen as further evidence that the resulting weak neural response might contribute to an overall destabilization of the cortical representation of stimuli paving the way for binocular rivalry's characteristic perceptual instability.
One way to evaluate this hypothesis is to investigate whether neural suppression is a precursor to all forms of binocular rivalry. If neural suppression does not occur under every circumstance of dichoptic stimulation that evokes binocular rivalry, it is unlikely to serve as a prerequisite for binocular rivalry. Here, we use binocular rivalry flash suppression (BRFS) to assess whether the ignition of binocular rivalry requires neural suppression at the moment of dichoptic onset. BRFS is a particular form of binocular rivalry where one eye is adapted before the onset of dichoptic stimulation. 79 This monocular adaptation period forces the secondarily presented -or flashed stimulus -to dominate perception for $1 s. 70,79,80 Importantly, the psychophysical properties of BRFS closely resemble those of binocular rivalry, suggesting that these two phenomena share a neural mechanism 29,41,73,79,81-83 (and see Ooi & Loop 84 ).
Our study shows that primate V1 neurons do not undergo unspecific response suppression at the onset of binocular rivalry when using BRFS. Instead, we observed transient response enhancement (facilitation). This outcome challenges the notion that response suppression is needed for initiation of binocular rivalry. We propose that the reduction of V1 activity due to adaptation may serve a similar role during BRFS. That is, adaptation affects both eye-specific and stimulus-specific V1 neural populations. This eye-specificity and stimulus-specificity of reduced responses may be one reason BRFS leads to a more deterministic perceptual outcome compared to regular binocular rivalry.

RESULTS
Our aim was to test the hypothesis that response suppression as a result of dichoptic stimulation functions as a prerequisite step for the initiation of binocular rivalry. To do this, we evaluated whether V1 spiking responses to BRFS diminished at the onset of interocular conflict (rivalrous stimulation) ( Figure 1). As a control, we first aimed to reproduce the existence of response suppression for conventional binocular rivalry that occurs following simultaneous onset of dichoptic gratings (Figures 2A and 2B). This dichoptic control allowed us to quantify the impact of interocular conflict on V1 spiking. Secondarily, we examined the influence of adaptation due to asynchronous onset on V1 responses during BRFS, because it is stimulus onset asynchrony (SOA) that differentiates BRFS from conventional binocular rivalry ( Figure 2C). To do so, we replicated the BRFS stimulus sequence, but presented a stimulus to the unadapted eye that was identical to (congruent with) the monocular adapter. This dioptic control allowed us to quantify the effects of BRFSrelated adaptation. As such, we were able to look for independent (dissociated) influences from adaptation and dichoptic conflict on V1 response rates.

Response suppression follows the onset of dichoptic stimulation
Prior studies on conventional binocular rivalry -with simultaneous onset of each eye's stimulus -found a reduction in V1 spiking responses. This response suppression occurred independently of the neurons' selectivity for stimulus orientation or the perceptual state of the subject (''interocular suppression'' or ''dichoptic cross-orientation suppression''). 61 Figure 1. Binocular rivalry flash suppression (BRFS) paradigm (A) Linear multielectrode arrays recorded neurophysiology data from primary visual cortex (V1) of two awake and behaving macaques (Macaca radiata). The cortical boundaries of V1 were determined with convergent results from local field potentials (LFP), current source density (CSD), power spectral density (PSD), mean multi-unit activity (MUA), and receptive field mapping (see STAR Methods). The granular input layer (4c) is the main reference point observed in LFP, CSD, and PSD. As such, layer 4c is set to a depth of 0mm, and the rest of the cortical depth is referenced from its lower laminar boundary. The electrode corresponding to -1mm below layer 4c does demonstrate significant responses to any receptive field, indicating that it is placed in subcortical white matter. (B) Eye preference and orientation tuning were determined for each V1 multi-unit. To be included in our analysis multiunits had to exhibit statistically significant response differences for eye stimulation and stimulus orientation.
(C) Subjects sat in front of a mirror stereoscope that allowed for monocular stimulation while the subject maintained binocular fusion. The subjects' task was to maintain fixation within a 1-degree window until the end of stimulus presentation. During BRFS trials, monocular adaptation for 800ms causes perception to switch to the unadapted eye when binocular rivalry begins at the onset of dichoptic presentation. Thought  iScience Article is typically observed 150-250ms after stimulus onset, which temporally matches the perceptual onset of binocular rivalry. 3,9,61,73,90 For full evaluation, we compared the responses of monocular stimulation (black) to both dichoptic (red) and dioptic stimulation (blue). Note that a preferred and non-preferred stimulus were chosen for each acute iScience Article penetration based on columnar population responses. Therefore, not every multi-unit on a given penetration had perfectly matching orientation and eye selectivity. We thus limited our statistical analyses to 91 (out of 219 tuned) V1 multi units that exhibited significant response preferences that aligned with our stimuli. Pairwise comparisons between conditions were evaluated for what we defined as both the transient (50-150ms) and sustained (150-250ms) response periods.
Dioptic stimulation initially resulted in higher multi-unit activity (MUA) responses than monocular stimulation during the transient phase of the visually evoked spiking response ( Figure 2A) 26 and Cox et al. 61 for further work on this subject).
Previous studies evaluating the suppression of neural responses as a result of dichoptic stimulation used a variety of stimulus contrast levels. 64,65,77,88,91,92 Here, we settled on medium-high contrast, balanced across the eyes (see STAR Methods). In order to rule out that our choice of contrast confounded the presence of response suppression during BRFS, we first tested for response suppression using our stimulus set. In other words, we first exposed the animals to regular binocular rivalry, where stimuli are presented simultaneously ( Figure 2B) to replicate the classic observation of response suppression as a result of dichoptic stimulation.
To do so, we simultaneously presented two dichoptic gratings at the same location of visual space, one in each eye (see STAR Methods for details). These dichoptic gratings had orthogonal orientations. 65,77 The initial transient response rate did not statistically differ between dichoptic stimulation and monocular stimulation. . Medium effect size is given by the matched rank biserial correlation (MRBC = À0.53). While constituting a replication of earlier findings, this response suppression is remarkable. It amounts to a single eye's stimulation eliciting more V1 excitation than stimulating both eyes simultaneously, despite retaining the same stimulus in the same eye.

Monocular adaptation greatly reduces the initial transient of binocular responses
We next revisited how V1 neurons respond to (monocular) adaptation. It is well established that adaptation elicits profound reduction of V1 responses.  In this specific case, we were interested in how far adaptation to one eye affects V1 responses to stimuli shown to both eyes. Figure 2C shows the effect of monocular adaptation on subsequent dioptic stimulation (i.e., how exposing one eye to a stimulus affects responses to adding the same stimulus to the other eye). This is an essential control for BRFS, as it has the same temporal and ocular stimulation pattern, modulo dichoptic conflict. As a result, this stimulation sequence causes perceptual fusion rather than binocular rivalry. Adaptation was always restricted to the neurons' non-dominant eye. After 800ms stimulus onset asynchrony, the second stimulus was shown to the dominant eye while the (adaptor) stimulus in the non-dominant eye remained on screen.
We found that monocular adaptation significantly reduced binocular neural responses during both the initial transient and sustained response windows. iScience Article MRBC = 0.889). To sum, monocular adaptation significantly reduces V1 responses to the onset of a dioptic stimulus (consisting of two copies of the adapter). It seems reasonable to assume that at least part of this response reduction also carries over to the situation where binocular stimulation consists of one copy of the adapter, paired with a novel stimulus (as is the case for BRFS). Since these results confirm that both (1) monocular adaptation reduces V1 responses and (2) dichoptic stimulation reduces V1 responses, we expected V1 responses to show a clear reduction at the onset of the dichoptic stimuli following monocular adaptation during BRFS.
Facilitation, not suppression, follows onset of BRFS As our final step, we investigated whether BRFS results in response suppression. To find out, we contrasted the onset of dichoptic stimulation during BRFS (light red in Figure 3) with the onset of dioptic stimulation following monocular adaptation (light blue in Figure 3). Prior work often described dichoptic response suppression with respect to monocular stimulation. 61,63,65 However, we chose to compare dichoptic to dioptic stimulation to account for the substantial neural suppression induced by monocular adaptation discussed iScience Article above. This shift in definition can be justified by the observation that sustained responses do not deviate between dioptic and monocular stimulation ( Figure 2A). In line with our revised definition, prior literature has consistently noted that dichoptic stimulation not only results in reduced responses when compared to monocular stimulation, but also when compared to dioptic stimulation. 61,[63][64][65]77,117 We first tested for decreased neural responses during the sustained response (150-250ms), where dichoptic response suppression was previously reported. 61 Note that this response period roughly coincides with a distinct phase during binocular rivalry onset where initial fusion subsides. 50 Dichoptic facilitation during BRFS is consistent across varying levels of excitatory drive So far, our considerations were limited to a specific combination of stimulus orientation and ocular configuration (i.e., preferred stimulus in dominant eye). We next widened our analyses to other combinatorial possibilities, as seen in Figure 4. This is an important comparison since it can help account for neural ''drive''. The concept of drive is equivalent to stimulus properties that enhance responses. Maximal drive was elicited Across these varying levels of drive, the lowest response was consistently evoked by dioptic presentations following monocular adaptation (Figure 4). A two-way repeated measures ANOVA was performed across the transient phase of all responses. The first factor is visualized as the black, blue, and red-indicated conditions in Figure 4. The second factor consisted of the stimulus type: preferred stimulus, non-preferred stimulus, preferred eye, and non-dominant eye. This factor is visualized as the groupings along the abscissa of Figure 4. Mauchly's test of sphericity indicates that the assumption of sphericity is violated (p < 0.05) in all samples, so a Greenhouse-Geisser sphericity correction was used. The two-way repeated measures ANOVA revealed a statistically significant interaction between the effects of condition type and stimulus presentation (F 3.33, 113 = 4.550, p = 0.003, h 2 = 0.041) (Tables 1 and 2). One way to interpret the above findings is that BRFS results in stronger responses than its dioptic counterpart since the dioptic control doubles presentation of an adapted stimulus, whereas BRFS introduces a novel, unadapted stimulus. This unadapted stimulus excites a largely unadapted population of V1 neurons.

DISCUSSION
BRFS 79,119 and its sister phenomena -continuous flash suppression (CFS) 120 and generalized flash suppression (GFS) 80 -are all closely related to binocular rivalry. Just like binocular rivalry, flash suppression paradigms utilize discrepant binocular stimulation to evoke suppression of one eye's view over the other (although there are also instances where interocular conflict is not a requirement). 8,80,81,121-124 And, just like binocular rivalry, prolonged stimulation under these paradigms leads to stochastic alternations, or reversals, between each eye's perspectives.

The role of primary visual cortex (V1) for binocular rivalry
One of the most widely debated questions about the neural basis of binocular rivalry is the role of V1. 17 Initial studies were confounded by the fact that fMRI was used in humans while measurements of spikes were performed in animal species. However, a more recent study showed that fMRI and single neuron spiking measurements in the same individuals still show this discrepancy 49 (see Anenberg et al. 138 for a possible mechanistic explanation, and see Watanabe et al. 72 for a different interpretation). In other words, the answer to whether or not perceptual alternations during rivalry strongly modulate V1 activity depends on the technique that is used to assess this question, not the specific model or paradigm. 31 To date, it is still unclear what causes this divergence between signals. Several more studies found divergences between iScience Article fMRI and local neural activity outside binocular rivalry, [139][140][141][142][143][144][145] suggesting that these two measurements are not as closely linked as previously assumed.
One potential explanation for this dissociation is that fMRI is more sensitive to subthreshold synaptic activation. [146][147][148][149][150] This hypothesis is supported by the finding that local field potentials (LFPs), which are believed to arise from (subthreshold) synaptic activity more closely predict the fMRI signal than local (suprathreshold) neuronal spiking. 148,151,152 Indeed, V1 LFP reflects perceptual changes during rivalry, in concert with the fMRI signal, in V1. 49 Feedback projections from higher visual areas to V1 are one potential source that might produce such subthreshold synaptic activity. 153 On this view, the V1 feedback modulation observed in fMRI and LFP is not fully translated into V1's spiking output, and is thus largely epiphenomenal in nature.
Another possibility is that the sampling of V1 neurons using standard in vivo neurophysiological techniques is systematically biased toward certain cell types and thus not fully representative of the whole neuronal population. In other words, fMRI might represent a population signal that weighs some types of neurons more heavily than others. There are several plausible possibilities for this to be the case, such as neurons in different layers of V1 differentially affecting the neurovascular coupling that underlies the fMRI signal. [154][155][156][157][158] On the flip side, it is also possible that there are different biochemical or morphological classes of neurons that are either under-or oversampled in in vivo neurophysiology. Some of the undersampled types of neurons might hold a disproportionate influence over neurovascular coupling and the hemodynamic response. 159,160 In other words, the relationship between the fMRI signal and local spiking may be (highly) non-linear and more complex with respect to specific populations of neurons. 161,162 Our study explicitly circumvents these issues by focusing on a simpler question: Does systemic suppression of V1 population activity at the initiation of rivalry trigger perceptual suppression of one eye's view (i.e., binocular rivalry)? Notably, this time window of interest precedes perceptual alternations of regular binocular rivalry. We asked whether the starting point for the neural mechanisms that result in perceptual limitation to one eye's view (perceptual suppression) can be found within V1's response. In other words, we are interested in the brief moment of onset rivalry, 50 and the neural ignition that precedes it.

V1 response suppression as a potential trigger mechanism for initiating perceptual suppression
What evidence speaks of V1 as the initiator for the onset of binocular rivalry? There are at least two lines of reasoning that justify this assumption. First, there is direct evidence derived from fMRI studies of BRFS, Second, the phenomenon of dichoptic (cross orientation) suppression that has been described on the level of single V1 neurons has already been implicated by researchers in the field. We will discuss each of these arguments in succession.
The first line of evidence is based on fMRI studies and stems from a modified BRFS paradigm. The most common set of stimuli for binocular rivalry and BRFS are (Cartesian) orthogonal gratings. Using these stimuli, perceptual dominance and suppression can occur in a piecemeal fashion, consisting of a mosaic of each eye's view. 2,163 Piecemeal rivalry is highly dynamic, giving the impression of fluid changes of the stimulus mosaic. Using non-Cartesian, radial stimuli (i.e., a sunburst pattern that rivals concentric rings), this fluidity becomes more ordered in that the transitions between each eye's view progress along the lines of the concentric rings. In other words, using these stimuli, perceptual alternations take on the form of traveling waves along a circular path.
Remarkably, this spatiotemporal pattern of a binocular rivalry traveling wave can be traced along V1's surface using fMRI, 47 as well as using voltage sensitive dyes in the primate model. 68 This stimulus can be used iScience Article for a locally spreading variant of BRFS. If the experimenter enhances the contrast of a small region of the stimulus in the perceptually suppressed eye, this ''flashed'' region will immediately gain perceptual dominance, just like in regular BRFS. Immediately following this local flash suppression, perceptual dominance continues to spread as a traveling wave. In other words, using this local variant of flash suppression, it is possible to trigger binocular rivalry traveling waves in V1, which are traceable using fMRI.
Using the above-mentioned visual paradigm in combination with an attention task yielded a surprising result: When an observer's attention is allocated to a part of visual space that does not contain interocular conflict, such as a fixation point in the center of the screen, flash-initiated binocular rivalry traveling waves are still observable with fMRI of V1. 132 And yet, these waves fail to be observed in all downstream visual areas. That is, when observers do not actively attend to the interocular conflict, V1 still initiates binocular rivalry, but the rest of the brain ignores that. One way to interpret this finding is that attention is required to ''gate'' V1's rivalry initiation signal to the rest of the visual system. 30, 164 Indeed, absence of attention seems to severely impact the perceptual outcome of both BRFS 165 and binocular rivalry. 69,72,166 Another interpretation is that V1 initiates perceptual dominance and suppression, with higher visual areas carrying out the actual task of establishing neural correlates of conscious perception. 167 This hypothesis of V1 serving as the locus of an initiator for binocular rivalry also rests well with the fact that V1 is the primary locus of binocular combination. 168,169 The second line of evidence stems from neurophysiological recordings in cat area A17/A18 (the homolog of primate V1). Recently replicated in macaques, these recordings demonstrated that binocular rivalryinducing stimuli lead to response suppression. 61,62,64,65,77,170,171 In theory, this widespread response suppression could serve as a trigger mechanism that allows V1 to prepare the rest of the visual system to induce alternating perceptual dominance and suppression.
One way to conceptualize this hypothesis is from the perspective of noise-driven attractor models of binocular rivalry. 75,172 A strong binocular response would resemble a stable attractor in the form of a global minimum. That is, whenever V1 faces stimuli that can be fused across the eyes, the resulting response is more stable than that of showing a stimulus to one eye only -far away from the random noise fluctuations that occur in the absence of visual stimulation. Such a stable attractor can be conceptualized as a deep cavity, or well, in a flat landscape that a ball rolls into. Once the ball enters such a deep well, it remains in position and the system keeps a steady state (hence: global minimum). When conflict is detected between the eyes, V1 responses are generally suppressed. This weakening of responses can be conceptualized as elimination of a global minimum that guarantees a stable state. Instead, V1's activity now is closer to (and thus more subject to) random noise fluctuations. As a result, V1's state is less stable. Using the attractor well metaphor again, we can assume that this means that the ball we envisioned is now not falling into a deep well anymore, where it will rest. Instead, smaller, local minima in the form of minute indentations in the flat landscape become new, but less reliable attractor states. That is, the random fluctuations between two semistable states (i.e., left eye's view versus right eye's view) are akin to two neighboring, small wells between which the system (ball) irregulary fluctuates. Once the ball falls into one of these two local minima, it will remain there for a limited amount of time. A system with constant energy flux (such as a ball remaining in motion) will eventually leave such a local minimum, or attractor state, only to get trapped again by the opposing, shallow well. All the above provides a metaphorical explanation for V1 as a potential locus to initiate binocular rivalry perception, with response suppression as a plausible trigger mechanism.
Response suppression from interocular conflict does not trigger perceptual suppression during BRFS At first glance, this finding seems incompatible with the idea that the initial stage of binocular rivalry is due to global destabilization of V1 activity via a reduced signal-to-noise ratio. However, as outlined below, this result is less surprising if interpreted through the lens of visual adaptation. Below, we discuss how our finding rules out interocular conflict induced response suppression as a generalized starting point of perceptual suppression for binocularly competing stimuli, it is still in line with many of the conceptual models that were based on that assumption.

Monocular adaptation as a mechanistic factor
BRFS is inherently an adaptation paradigm. Monocular adaptation is required to occur for an absolute minimum of 500ms for effective flash suppression to occur. 29, 79 The main finding of our study is a surprising facilitation of the visual transient following the rivalrous onset of BRFS. This transient BRFS response facilitation can be explained by assuming distinct populations of V1 neurons for each eye and stimulus combination. For example, presenting a horizontally oriented grating to the left eye should maximally excite neurons that prefer the left over the right eye as well as horizontal gratings over vertical gratings. The rivaling stimulus, consisting of a vertically oriented grating presented to the right eye, should maximally excite a non-overlapping population of V1 cells (right eye and vertical grating preferring neurons).
In this scenario, the initial phase of BRFS should maximally adapt (fatigue) one set of neurons. When the second stimulus is presented (flashed) to the other eye, the flashed stimulus will maximally excite a population of neurons that has not been adapted. Indeed, we found that following 800ms of adaptation, V1 responses to the flashed stimulus were nearly identical to those of a simple monocular onset. It is noteworthy that even non-preferred BRFS stimuli evoked neural responses above those of their dioptic counterparts. Yet, this is not surprising since these conditions, too, resemble presentation of a novel stimulus in the first condition and a repeated stimulation with an adapted stimulus in the latter case. In other words, adaptation seems to cause the reduction in responsivity that enables the heightened BRFS transient response.
In this context, it is worth pointing out that this automatic -and thus trivial explanation -for an increased neuronal response to the flashed stimulus (which dominates perception) was already noted by preceding studies on the neural correlates of consciousness that used BRFS. For this reason, these studies omitted the initial transient from their analyses. 28,29,173 Thus, the finding of neuronal correlates of conscious perception reported in these studies is not affected by the mechanistic account of adaptation presented here.
Interestingly, recent work in awake behaving primates demonstrated that there are suppressive interactions between each eye's signals as early as in the LGN, as well as the input layers of V1. 26,116,169 Similarly, interocular inhibitory activity was also reported by studies of interocular transfer of adaptation. 21,73,171,[174][175][176][177] In other words, left eye and horizontal grating preferring neuronal populations as well as right eye and vertical grating preferring neurons are expected to not only co-exist as two neural populations with nearly orthogonal response properties, but they are also likely to exhibit mutually suppressive interactions. 21,33,56,60,73 Assuming this mechanism to exist, flashing the second stimulus not only excites an unadapted population of cells, but also releases their inhibition, thus further explaining the neural facilitation observed in our paradigm.
Lastly, it is worth noting that we observed some form of interocular conflict induced response suppression in the case of BRFS. However, the timing of this interocular conflict induced response suppression is significantly later than previous characterizations, 61 excluding interocular conflict induced response suppression from having causal relevance for the initiation of perceptual suppression. In other words, the interocular conflict induced response suppression that accompanies BRFS occurs after the onset of perceptual suppression that characterizes this paradigm. 79 It is nonetheless noteworthy that response suppression as a result of dichoptic stimulation is a ubiquitous phenomenon in V1. Recent work demonstrated that a suppressive response to binocular stimuli is not exclusive to (dichoptic) stimuli that cannot be fused, but even occurs for some matching (dioptic) stimuli that do not result in binocular rivalry. 26 Binocular fusion is arguably one of the main functions of the early visual system, as it gives rise to stereopsis. 178 Interocular conflict is constantly experienced because of nasal occlusion, 173  iScience Article and non-foveal monocular representations. As such, binocular rivalry is a common occurrence in natural viewing. 179 Interocular conflict detection and binocular fusion are rapid and automatic processes that occur with every eye movement or blink in natural viewing. Within this context, BRFS is a useful paradigm, as the flashed dichoptic onset mimics interocular discrepancies following a saccade without the confounds of eye movement related activity.

Adaptation might serve as an alternate trigger for BRFS
If neural response suppression does not induce perceptual suppression during BRFS, what mechanism might be responsible? One interesting clue is that adaptation generally induces reduced V1 neural activity. 102,116 Thus, while interocular conflict induced response suppression does not seem to be at play during BRFS, reduced V1 activity could still be a factor.
In contrast to the global effect of interocular conflict induced response suppression, which affects all stimulus-responsive V1 neurons, the reductive effects of adaptation are more specific. Above, we assumed that there are two non-overlapping populations of V1 neurons targeted by a BRFS stimulus, and we assumed that one population is fatigued during the adaptation phase of BRFS. However, we excluded a large fraction of V1 neurons that does not fall into either of these two distinct groups: The majority of V1 neurons receive inputs from and respond to from both eyes. 24 These binocular neurons also undergo adaptative suppression thanks to interocular transfer of adaptation. 21,73,171,[174][175][176][177] As a result, the adaptive period of BRFS results in rather wide (albeit not global) suppression of V1 responses for the stimulated region of the visual field. As a result, noise-driven attractor models could still function within the context of BRFS given that response suppression as a result of dichoptic stimulation was replaced with suppression from adaptation. As such, adaptation might be the mechanism that facilitates the onset of perceptual suppression in BRFS via reduction of visual responses.
We speculate that a similar mechanism might drive the psychophysical dynamics of continuous flash suppression (CFS). 71,82,120,134,180,181 During CFS, Mondrian patches are consistently presented to one eye, thereby continually suppressing the image in the other eye from obtaining perceptual dominance. Thus, BRFS and CFS might operate on a similar mechanism. Each novel Mondrian patch flash captures previously unadapted neuronal populations, and thus continually promotes that eye's view into perceptual awareness. 82 It is also intriguing to speculate that we need to look beyond rate codes to fully understand and describe V1's role for the initiation and maintenance of binocular rivalry. In neural systems, signal flow in the form of increased spiking responses and information flow are fundamentally dissociable. 6,182 There are several well-known causally relevant phenomena that occur exclusively at the population level. One example of such phenomena is neuronal synchrony, 13,183,184 while another is systematic changes in correlated noise. 185,186 Neither of these causal computations can be detected in analyses of response magnitude alone. Indeed, binocular rivalry has been shown to evoke changes in neuronal coherence in both cats 187 and primates. 38 Given what we stated above about the potential of functional differences across cortical layers, it would be intriguing to apply such population-based measures with laminar resolution for the study of binocular rivalry in V1.
Conclusion: What is the role of adaptation for binocular rivalry more generally?
Despite inclusion of adaptation-related parameters in several newer models of binocular rivalry, 56,60 the influence of adaptation on the neurophysiological concomitants of binocular rivalry remains largely unknown. 21 There are several reasons for an underdeveloped understanding of adaptation's influence on neural responses during binocular rivalry. First, adaptation-based opponency models that lack an ad hoc source for random noise demonstrate rivalry oscillations produce regular intervals for perceptual alternations, which is an inaccurate reflection of the binocular rivalry process 75,188 (and see Cha & Blake. 189 ).
Noise-driven attractor models improve upon adaptation-based opponency models by accounting for random fluctuations in perception, 75  iScience Article Limitations of the study One notable caveat with our interpretation is that in the initial documentation of response suppression as a result of dichoptic stimulation in anesthetized cats, the phenomenon was evaluated for several seconds after stimulus onset. [63][64][65]77,117 Later investigations in macaques describe suppression 150ms-250ms after stimulus onset, 61 leaving some ambiguity about the exact timing of this phenomenon.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: Multi-unit responses were taken from the MUA signals described above. Each session was trial-averaged per stimulus condition and then z-score normalized. The z-score normalization was relative to the pre-stimulus baseline responses for each electrode contact. Baseline responses were determined as the mean response for the 50ms leading up to but not including stimulus onset while the subject was fixating on the mean-luminance grey screen. Baseline averages and standard deviations were calculated for all baseline periods over the recording session. The mean and standard error on the mean of condition-specific responses were taken for all contacts within cortex.
Multi-units were classified by their stimulus preferences. To do so, responses were averaged over the 50ms-250ms time-period after stimulus onset for the monocular presentations varying in eye, orientation, contrast, spatial frequency, and phase. Eye and orientation tuning was obtained with MATLAB's anovan.m function to see if a unit's response rates varied with both eye-of-origin and orientation. For a unit to be included in all subsequent analyses, it had to be significantly tuned to both eye-of-origin and orientation.
Each unit's specific orientation preference was determined with a Gaussian curve fitted to the mean response profiles for each sampled stimulus orientation. 18 different gratings of 10⁰ increments from 0⁰ to 180⁰ were used. Preferred orientation was determined by obtaining the maximum of the curve. Null orientation was set as a 90 o rotation from the preferred orientation.
We used characteristics of local field potentials (LFP), current source density (CSD), power spectral density (PSD), and multi-unit activity (MUA) to determine cortical boundaries, 6,24,61,116 see 58 for histological verification. Additionally, we ensured receptive fields were aligned across cortical depths. Aligned receptive fields indicate that our acute electrode penetrations were orthogonal to the cortical surface, and aid in the determination of V1 upper and lower boundaries. Out of 360 recordings (24 contacts on each laminar probe, used across 15 sessions), 219 were determined to be within V1 cortical boundaries (142 in ''E'' and 77 in ''I''). Of these 219 V1 multi-units, 128 did not show significant tuning to both eye and orientation and were removed from the analysis. 91 V1 multi-units demonstrated significant tuning to both eye and orientation (72 from ''E'', and 19 from ''I''), and were used in all subsequent analyses. Table 2 shows the subject and sample information.
Statistical analyses were performed via MATLAB or using the open-source statistics JASP software package (JASP Team (2022). Version 0.16.3). When assumptions of normality were violated, non-parametric tests were used. Our data analysis primarily comprised of pairwise comparisons, so mean MUA responses were evaluated via Wilcoxon signed-rank tests.

ll
OPEN ACCESS