Open Access
Article  |   November 2019
Center-surround velocity-based segmentation: Speed, eccentricity, and timing of visual stimuli interact to determine interocular dominance
Author Affiliations
  • Egor Ananyev
    Nanyang Technological University, Department of Psychology, Singapore
    egor.ananyev@gmail.com
  • Zixin Yong
    Duke-NUS Medical School, Neuroscience and Behavioural Disorders Program, Singapore
  • Po-Jang Hsieh
    National Taiwan University, Department of Psychology, Taipei, Taiwan
    hsiehpj@ntu.edu.tw
Journal of Vision November 2019, Vol.19, 3. doi:https://doi.org/10.1167/19.13.3
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Egor Ananyev, Zixin Yong, Po-Jang Hsieh; Center-surround velocity-based segmentation: Speed, eccentricity, and timing of visual stimuli interact to determine interocular dominance. Journal of Vision 2019;19(13):3. https://doi.org/10.1167/19.13.3.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

We used a novel method to capture the spatial dominance pattern of competing motion fields at rivalry onset. When rivaling velocities were different, the participants reported center-surround segmentation: The slower stimuli often dominated in the center while faster motion persisted along the borders. The size of the central static/slow field scaled with the stimulus size. The central dominance was time-locked to the static stimulus onset but was disrupted if the dynamic stimulus was presented later. We then used the same stimuli as masks in an interocular suppression paradigm. The local suppression strengths were probed with targets at different eccentricities. Consistent with the center-surround segmentation, target speed and location interacted with mask velocities. Specifically, suppression power of the slower masks was nonhomogenous with eccentricity, providing a potential explanation for center-surround velocity-based segmentation. This interaction of speed, eccentricity, and timing has implications for motion processing and interocular suppression. The influence of different masks on which target features get suppressed predicts that some “unconscious effects” are not generalizable across masks and, thus, need to be replicated under various masking conditions.

Introduction
Binocular rivalry occurs when conflicting stimuli are presented to the two eyes, which leads to continuous switches of perception from one stimulus to the other. Rivalry presents an opportunity to study perceptual awareness (Blake & Logothetis, 2002; Crick, 1996; Crick & Koch, 1998; Giles, Lau, & Odegaard, 2016). There is substantial evidence that conscious perception of a stimulus—and the suppression of the other—is a culmination of a chain of hierarchical decisions by the perceptual system (Blake & Logothetis, 2002; Logothetis, Leopold, & Sheinberg, 1996). The basis for such perceptual decisions has been psychophysically evaluated by considering binocular rivalry in conjunction with visual features, such as spatial frequency (Blake & Fox, 1974; Nguyen, Freeman, & Wenderoth, 2001; O'Shea & Crassini, 1981; Stuit, Cass, Paffen, & Alais, 2009; Walker & Powell, 1979), color (Carlson & He, 2000; Fox & Check, 1968; Hong & Blake, 2009; Kovács, Papathomas, Yang, & Fehér, 1996; Livingstone & Hubel, 1987; Nguyen et al., 2001; Ooi & Loop, 1994; O'Shea & Williams, 1996; Smith, Levi, Harwerth, & White, 1982), and size (Blake, Yu, Lokey, & Norman, 1998; Blake, Zimba, & Williams, 1985). Binocular rivalry of moving stimuli was also investigated (Blake et al., 1998). Dominance in binocular rivalry was shown to depend on velocity with a moving stimulus typically dominating over a static stimulus (Blake et al., 1998; Blake et al., 1985; Breese, 1899; Wade, Weert, & Swanston, 1984; Wiesenfelder & Blake, 1990). This affinity toward dynamic stimuli was utilized by interocular suppression techniques, such as continuous flash suppression (Tsuchiya & Koch, 2005). Flashing stimuli presented to one eye effectively mask a target stimulus presented to the other eye for long durations (Stein, Hebart, & Sterzer, 2011; Yang, Brascamp, Kang, & Blake, 2014). With the popularity of such interocular dynamic suppression methods, it is important to establish the mechanisms of dominance and suppression of motion. In addition, examining binocular rivalry in conjunction with motion perception may elucidate the neural mechanisms involved in both processes. 
Both binocular rivalry and motion processing depend on size and eccentricity of the stimuli. A larger foveally presented stimulus extends farther into periphery; eccentricity is, therefore, intrinsically linked to the size of such stimulus. In case of binocular rivalry, larger stimuli were associated with so-called mixed (piecemeal) dominance, and smaller stimuli often resulted in whole-field dominance (Blake, O'Shea, & Mueller, 1992). Similar to a decrease in size, presenting stimuli in the periphery resulted in whole-field dominance increase. This was accounted for by the increase in the receptive field size with further eccentricities (Cowey & Rolls, 1974). Centrally presented superimposed motion patterns can be readily perceptually individuated (De Bruyn, 1997). This is perceived as partial transparency of either stimulus. However, peripherally presented motion results in integrated (averaged) motion perception (De Bruyn, 1997). Likewise, the participant readily extracts motion direction of an object viewed through apertures in the periphery but not centrally (Lorenceau & Shiffrar, 1992). Furthermore, so-called “form constraints” on motion integration are weaker in the periphery (Lorenceau & Alais, 2001). Foveally, an occluded “closed” shape (diamond) results in coherent motion more often than an “open” shape (e.g., cross). In the periphery, however, motion integration accuracy was found to be high regardless of the occluded shape. These results suggest that peripheral presentation benefitted motion integration. 
The goal of the current study was to investigate the spatial aspect of motion rivalry. Specifically, we sought to probe the binocular dominance (Experiments 14) and suppression strength (Experiment 5) as a function of eccentricity. Given the effects of eccentricity on both binocular rivalry and motion processing, it is surprising to see no reports of mixed or piecemeal dominance of moving stimuli in previous studies of motion rivalry (Blake et al., 1985; Marshak & Sekuler, 1979; van de Grind, van Hof, van der Smagt, & Verstraten, 2001; Wiesenfelder & Blake, 1990). Part of the issue may be that the popular moving dot stimuli have form information that may alter the dynamics of rivalry (Alais & Parker, 2006, 2012). We, therefore, used a set of stimuli known as “motion clouds” to minimize form information (Leon, Vanzetta, Masson, & Perrinet, 2012). To capture the spatial dominance pattern of rivalry, we had participants evaluate it at a fixed point in time, e.g., at the end of a brief, 1-s-long trial. This approach is known as onset rivalry (Carter & Cavanagh, 2007; Stanley, Forte, Cavanagh, & Carter, 2011). To preview the results, we found that dominance was defined by velocity-based center-surround segmentation. Namely, the center and surround were dominated by the slower and faster stimuli, respectively. 
Experiment 1
Methods
Participants
Five participants, including the first author (S1), took part in the experiment. All signed an informed consent form, and all except the author were compensated for their time. For the purposes of showing individual data, the subject IDs (SX) were kept consistent for participants across experiments. The study was approved by the Duke-NUS institutional review board and adheres to the Declaration of Helsinki. 
Apparatus
Stimuli were presented against a black background on a 22-in. Samsung 2233RZ LCD monitor with a resolution of 1,680 × 1,050 pixels and a refresh rate of 60 Hz. The participants viewed the stimuli through a custom-made, four-mirror stereoscope mounted on a chin rest. The stimuli were presented and responses recorded using the PsychoPy stimulus presentation package for Python (Peirce, 2007, 2009) on a Macintosh. 
Stimuli
Stimuli (Figure 1) were naturalistic motion clouds, which are spatiotemporally filtered dynamic noise patterns (Leon et al., 2012). The stimuli were viewed through an aperture 6.4° of visual angle in diameter (except for Experiment 2, in which the size was varied). Stimuli of all sizes were enclosed within a 6.4° circle with a black edge. Spatial frequency bandwidth peaked at 3.5 c/° of visual angle (average spatial frequency = 1.75 c/°; see Supplementary Files S1S5 for stimulus examples). 
Figure 1
 
An example trial and trial structure. Motion clouds moving at seven different translational velocities (from 0°/s to 16°/s) in opposite directions were presented to two eyes for 1 s. The arrows inside the stimuli schematically show velocity and direction of motion. In this example trial, the stimulus presented to the right eye had a higher velocity. The example percept was that of the slower stimulus in the center and the faster stimulus along the edges (illustrated schematically in the upper-right corner). The participant was to indicate the dominance of the whole field or the central region only if the center differed from the surround. Because the percept (upper right) was center-surround, the participant indicates leftward dominance (i.e., dominance in the center). Feedback was provided with a leftward notch reflecting the response. To indicate that leftward motion was only dominant in the center, the participant adjusted the size of the red circle to approximately reflect the area of the center. In this case, the center size was indicated to be ∼75% of the stimulus diameter. Note that the faster stimulus (right) introduces motion streaks due to high velocity; these are eliminated in further experiments by using temporal frequency modulation instead of directional motion (see Supplementary Files S1S5 for stimulus examples).
Figure 1
 
An example trial and trial structure. Motion clouds moving at seven different translational velocities (from 0°/s to 16°/s) in opposite directions were presented to two eyes for 1 s. The arrows inside the stimuli schematically show velocity and direction of motion. In this example trial, the stimulus presented to the right eye had a higher velocity. The example percept was that of the slower stimulus in the center and the faster stimulus along the edges (illustrated schematically in the upper-right corner). The participant was to indicate the dominance of the whole field or the central region only if the center differed from the surround. Because the percept (upper right) was center-surround, the participant indicates leftward dominance (i.e., dominance in the center). Feedback was provided with a leftward notch reflecting the response. To indicate that leftward motion was only dominant in the center, the participant adjusted the size of the red circle to approximately reflect the area of the center. In this case, the center size was indicated to be ∼75% of the stimulus diameter. Note that the faster stimulus (right) introduces motion streaks due to high velocity; these are eliminated in further experiments by using temporal frequency modulation instead of directional motion (see Supplementary Files S1S5 for stimulus examples).
The fields were moving at one of seven velocities (0°/s, 0.5°/s, 1°/s, 2°/s, 4°/s, 8°/s, and 16°/s) in opposite directions along the x-axis except when the velocity was zero. This resulted in 48 velocity combinations (minus the 0–0 pair), which, balanced for each eye, led to a total of 96 trials per run. All stimuli had micromotion of 0.5°/s in bandwidth (i.e., had velocities from 0–0.5°/s). This was a directionally isotropic temporal frequency modulation (i.e., no net motion direction), which was independent of the directional (leftward/rightward) speed manipulation. For simplicity, we further refer to stimuli with zero translational velocity (i.e., only with micromotion) as stationary. For Experiment 2 and onward, this micromotion was eliminated for the stationary stimulus. At maximum velocity of 16.5°/s, the temporal frequency range corresponded to 0–57.6 c/s. 
Procedure
The trials were 1 s in duration. At the end of each trial, the participants were to characterize the percept right before trial termination. The progression was self-paced, and the participants were encouraged to rest during and between runs. 
The participant's task in each trial was to report the dominance. To do this, the participant first indicated the dominant motion and then the spatial pattern of that dominance. There were eight possible perceptual outcomes (Table 1). The dominance outcomes were fast, slow, stationary, or other (transparent/piecemeal). The spatial pattern of dominance was either whole-field or center-surround. The “center-surround” outcome referred to trials in which one stimulus dominated the center and the other dominated the surround (see Figure 1, upper right, for illustration). 
Table 1
 
The percentage of trials of different dominance judgments in terms of the velocity (column) and spatial distribution (rows). Notes: Trials with both velocities being equal (2.5% of trials) were not analyzed for responses. The standard deviation is in parentheses. The italicized cells indicate dominance categories omitted from further analysis due to low incidence (<5% of all trials).
Table 1
 
The percentage of trials of different dominance judgments in terms of the velocity (column) and spatial distribution (rows). Notes: Trials with both velocities being equal (2.5% of trials) were not analyzed for responses. The standard deviation is in parentheses. The italicized cells indicate dominance categories omitted from further analysis due to low incidence (<5% of all trials).
For fast or slow direction dominance, participants indicated the direction of the dominant motion by pressing the arrow keys (Figure 1). If the velocity in the center was different from the surround, the participants indicated the dominance of the center. In these center-surround cases, they adjusted a red circle approximately describing the area of the center (size judgment). The circle size could be varied in 10 steps by pressing up/down arrow keys. If the velocity was the same for the entire field, however, the participants were to leave the red circle at the maximum diameter setting, where it remained by default. 
If no directional motion was seen (“stationary” dominance), the participants pressed the “down” arrow key. For transparency or mixed rivalry, the participants pressed the “up” arrow key. The participants had the option of indicating a center-surround spatial pattern for these dominance outcomes as well. For example, if the participant thought that mixed rivalry was only in the center, they adjusted the circle to indicate its approximate size (“size judgment” above). Three participants (S1–3) completed three runs. Two additional participants (S4–5) completed two runs as this number of runs was deemed sufficient. Each run included 288 trials (96 trials per run × counterbalanced for 2 eyes) and took 10–20 minutes to complete. 
Top-down factors play a limited role in onset binocular rivalry (Dogge, Gayet, Custers, & Aarts, 2018). Nevertheless, care was taken to avoid any bias in instructions for each participant. A drawing was used to explain all potential percepts and responses. A center-surround example given involved fast motion in the center and stationary stimulus in the surround, i.e., the opposite of the pattern observed on most trials (stationary center/fast surround). The participants were encouraged to ask questions if they needed clarification. 
Analyses
All analyses were carried out in R statistical package (R Core Team, 2014). Mixed-effects linear models were used in statistical analysis with participant's intercept as a random factor, using functions from lme4 library (Bates, Mächler, Bolker, & Walker, 2015). Mixed-effects models were used due to high statistical power and resistance to unequal number of trials per condition (Baayen, Davidson, & Bates, 2008; Cohen, Cohen, West, & Aiken, 2003). Due to mathematical ambiguity in computing degrees of freedom for mixed models (Baayen et al., 2008), they were not provided. 
There were three dependent variables of interest: type of dominance (slow, fast, stationary, or other), proportion of trials with center-surround dominance, and the size of the center in the center-surround trials. The first two outcomes were analyzed with binomial generalized linear models (GLMs), so z values were provided; the size of center was analyzed with a GLM, so t values were provided. The independent variables were the competing speeds of motion in the two eyes. 
In addition to the linear models, the BayesFactor package was used to compute Bayes Factor (BF) values for significance testing (Morey, Rouder, & Jamil, 2014). Specifically, we used function lmBF to compare nested models including (nominator, H1) or excluding (denominator, H0) the variable of interest with participant as the random factor (Rouder & Morey, 2012). The utilized approach is known as the “default” BF as it does not require the specification of priors (Morey et al., 2014). BF values are a powerful alternative to P values as they can provide evidence toward the null hypothesis (i.e., distinguish between supporting the null hypothesis and insufficient data). Per convention (Jeffreys, 1939), BF-values below 1/3 were taken to support the null hypothesis (the denominator model), above three the alternative hypothesis (the nominator model), and the values in between indicating insufficient evidence to support either. 
The ggplot2 package was used to visualize the data (Wickham, 2009). 
The complete code of the experimental procedure and statistical analysis is available on https://github.com/egorananyev/mc
Results
Generally, if one stimulus was fast and the other slow, the space was segmented into a slow center and fast surround. If both stimuli were slow, this resulted in the whole-field slow pattern. As a result, slow stimuli dominated either the whole field or in the center of stimulus space on the majority of trials. At the same time, whole-field fast motion dominance was rare. 
We first looked at the overall proportion of stationary dominance outcomes (Table 1 and Figure 2A; see Supplementary Figure S1 for individual data). A stationary dominance outcome referred to trials when the dominant stimulus had a zero translational velocity or was too slow for the direction to be visible. Whole-field stationary stimulus dominance occurred on 34.1% of all trials. This occurred mostly on trials when both competing velocities were slow (b = −0.74, z = −11.13, p < 0.001, BF > 10125; Figure 2B). Additionally, 27.1% of all trials had the stationary stimulus dominant in the center with the faster stimulus dominant along the edges (“center-surround” stationary trials). Most of these stationary center dominance trials were associated with cases when one competing velocity was slow and the other fast (Figure 2C). We called this phenomenon “center-surround velocity-based segmentation.” Together, whole-field and center-surround stationary dominance accounted for >60% of all trials across all velocity combinations. 
Figure 2
 
Experiment 1: Trials reported as stationary-dominant or center-surround. (A) The proportion of stationary-dominant (both whole-field and center-surround) trials. Top: Stationary-dominant trial proportion plotted as a function of the lower (x-axis) and higher (y-axis) of the two competing velocities. Bottom: Same trials plotted as a function of the higher velocity (x-axis). When both velocities were sufficiently low, stationary dominance was most frequent (top graph: lower-left corner; bottom graph: left side of the plot). Interestingly, the slow stimulus also dominated on a majority of trials when competing with a high-velocity stimulus (top graph: upper-left corner; bottom graph: right side of the plot). When the higher velocity was medium (e.g., 4°/s in bottom graph), stationary dominance was less likely to occur. (B) The number of whole-field trials given stationary dominance. These trials occurred more often when both velocities were low (top graph: bottom rows). (C) The number of center-surround trials given stationary dominance. Such trials were closely associated with high-versus-low velocity competition (top graph: upper rows). Dots in the top graphs are group averages. Bottom graphs show standard error (N = 5).
Figure 2
 
Experiment 1: Trials reported as stationary-dominant or center-surround. (A) The proportion of stationary-dominant (both whole-field and center-surround) trials. Top: Stationary-dominant trial proportion plotted as a function of the lower (x-axis) and higher (y-axis) of the two competing velocities. Bottom: Same trials plotted as a function of the higher velocity (x-axis). When both velocities were sufficiently low, stationary dominance was most frequent (top graph: lower-left corner; bottom graph: left side of the plot). Interestingly, the slow stimulus also dominated on a majority of trials when competing with a high-velocity stimulus (top graph: upper-left corner; bottom graph: right side of the plot). When the higher velocity was medium (e.g., 4°/s in bottom graph), stationary dominance was less likely to occur. (B) The number of whole-field trials given stationary dominance. These trials occurred more often when both velocities were low (top graph: bottom rows). (C) The number of center-surround trials given stationary dominance. Such trials were closely associated with high-versus-low velocity competition (top graph: upper rows). Dots in the top graphs are group averages. Bottom graphs show standard error (N = 5).
The faster of the two competing stimuli dominated the whole field on only 9.5% of all trials (Table 1). Whenever the direction of the slow motion was visible (i.e., the participant indicated the dominant direction of motion), the percentage of slow-dominant whole-field trials was comparable to fast-dominant trials (9.7%; Table 1 and Supplementary Figure S2). 
Due to low incidence of center-surround patterns with fast-/slow-direction dominance (2.6% each) and other dominance outcomes (piecemeal/transparent center-surround, 0.8%), these trials were omitted from further analysis (the grayed-out cells in Table 1). 
The most prominent outcome was “stationary” dominance, either in the center or the whole field. It occurred when the dominant translational velocity was either zero or too slow to be visible. This perceptual outcome was more common in trials when the slower of the two competing stimuli was itself slow (b = −1.07, z = −7.16, p < 0.001; BF > 1042; Figure 2A). In other words, when one of the competing stimuli was slow, it was likely to dominate either in the center or the whole field. The effect of the higher velocity was not significant, probably due to the bimodality of the distribution (p = −0.008; BF = 0.08; Figure 2A, bottom). Specifically, stationary dominance occurred when the competing speed was either low or high, but not for intermediate velocities, on both whole-field and center-surround trials. This effect was consistent across participants (Supplementary Figure S1). 
We next looked at the size judgment in the static-center/fast-surround trials. The size of the stationary “center” increased slightly with the increase in the higher velocity, but remained constant with no influence of the lower velocity (higher velocity: b = 0.009, t = 3.20, p = 0.001, BF = 4.86; neither lower velocity nor interaction effects were significant: p > 0.1, BF < 1/3; Supplementary Figure S2). 
In trials when the dominant direction was indicated (nonstationary-dominant trials), fast and slow motions were about equally likely to dominate (9.5% and 9.7%, respectively), mean difference = 0.007, paired t test t(20) = 0.82, p = 0.422 (Table 1). Both fast- and slow-direction dominance were primarily driven by intermediate velocities. This can be seen at the bottom of Figure 2A, where the stationary outcomes “dipped” when the higher velocity was intermediate (x-axis from 2°/s to 8°/s), meaning that, for those velocity pairs, the direction of dominant motion was indicated. When these “direction-dominant” trials were analyzed separately, the fast- and slow-direction trials had very similar velocity-dependence profiles (compare Supplemental Figure S3A and S3B). 
The occurrence of transparency or piecemeal rivalry (13.4% of trials) was associated with the competition of two high velocities (lower velocity effect: b = 0.723, z = 4.40, p < 0.001; BF > 1045; Supplemental Figure S1B); i.e., as the lower of the two velocities increased, the occurrence of piecemeal/transparent trials also increased. Although no distinction was made behaviorally, unstructured anecdotal descriptions primarily characterized high-velocity pairs as transparent, i.e., both motion stimuli seen as overlapping with each other. 
Discussion
A novel phenomenon was observed when two motion clouds of different velocities competed with each other for short duration: If one stimulus was stationary or slow and the other was fast, the stimulus space was segmented into a slower center and a faster surround on a majority of trials. The incidence of the center-surround pattern and the size of the center both depended on the speed of the faster stimulus: the higher velocities produced more center-surround trials with larger centers. 
Another surprising finding was the prevalence of whole-field stationary dominance, which accounted for more than a third of all speed combinations (34.1%; Table 1). For example, when 2°/s motion was the faster of the two competing stimuli, the resulting percept was stationary on 75% of the trials (Figure 2B). Even for intermediate velocities (4°/s–8°/s), the frequency of stationary dominance was nonnegligible. Because the motion in the two eyes was always in opposite direction, this may indicate some degree of cancelling of one motion by the other, resulting in a stationary percept. If so, this would be similar to an effect observed with binocular spiral motion, in which opposing spiral rotation led to motion cancellation (de Weert & Wade, 1984). 
The observed dominance of stationary percepts could potentially be explained by the difference in contrast sensitivity as a function of stimulus speed (Kelly, 1979). One of the consequences of increasing temporal frequency is a drop in contrast sensitivity, which could account for why the faster motion dominates in the periphery. The effective spatial frequency decreases with eccentricity due to lower acuity, resulting in lower perceived spatial frequency toward the edges of the stimuli. The lower the spatial frequency, the higher the contrast sensitivity to high-velocity stimuli (Kelly, 1979). As a result, fast motion dominates the periphery. Based on this theory, we can predict that a smaller stimulus should abolish the observed spatial segregation into slow center and fast surround. We tested this prediction in Experiment 2. To preview, this prediction was only partially confirmed. Another caveat is that, per Kelly's data, sensitivity for the spatial frequency used here (average sf = 1.75 c/°) should peak at ∼3°/s, yet such intermediate velocities rarely dominated in our case. 
Another prominent outcome was stimulus transparency. If both stimuli were fast, both movement directions were visible simultaneously with the appearance of semitransparency (Blake et al., 1985; Marshak & Sekuler, 1979; van de Grind et al., 2001). This result is consistent with the modeling work suggesting that higher divergence in motion signal leads to transparency (Medathati, Rankin, Meso, Kornprobst, & Masson, 2017). In terms of modeling the velocity response, higher velocity divergence between the two rivaling motion clouds should result in more defined velocity peaks. Medathati et al. (2017) show that, when the two motion signals cannot be reconciled (vector averaging) or result in a clear dominance (winner takes all), the two motion carriers are seen simultaneously as spatially overlapping, resulting in transparency. 
Having translational velocity allowed us to test motion combinations in which both stimuli were nonstationary. However, one shortcoming was that the greater velocities in Experiment 1 produced horizontal motion streaks that artificially introduced orientation information (Apthorp, 2011; Apthorp & Alais, 2009). They can be seen as the horizontal component in the right eye stimulus in Figure 1 and could be interpreted as contamination of motion by “form” information. To reduce the role of potential confounds (form and direction), in Experiments 24, we eliminated translational velocity in favor of velocity bandwidth. 
Experiment 2
Our next objective was to investigate the spatiotemporal characteristics of the central bias toward completely stationary stimuli (i.e., without micromotion). We investigated the effects of size (Experiment 2), relative stimulus onset time (Experiment 3), and trial duration (Experiment 4). We also switched from directional velocity to bandwidth-based velocity manipulation. In particular, motion clouds were either stationary or fast, i.e., 0°/s or 16°/s peaks in speed bandwidth, respectively. This setup eliminated motion streaks and simplified behavioral responses for the participants. 
The objective of Experiment 2 was to establish the role of size in the center-surround velocity-based segmentation. The manipulation of the stimulus size was shorthand for the effect of eccentricity as a larger stimulus stimulates farther eccentricities. Consequently, we varied the sizes of both competing stimuli. If the stationary center dominance were tied to foveal or near-foveal vision, presenting a smaller stimulus would result in a whole-field stationary percept and would confirm the hypothesis that the slow center/fast surround segmentation is the product of increased sensitivity to high velocity in the periphery (Kelly, 1979). If, however, the stationary center scaled to the stimulus size, we may conclude that the stationary-center/fast-surround percept has a scale-invariant component. 
Methods
The same participants took part in this experiment as in Experiment 1 except the first author (S1) was replaced by an additional participant (S6). Participants S2–4 completed two runs; participants S5 and S6 completed only one run. The methods were also the same as in Experiment 1 with the following exceptions. 
Stimulus speed was set to velocity bandwidth that peaked at either 0°/s (stationary without micromotion) or 16°/s (dynamic). On each trial, stimulus size was pseudorandomly assigned to one of six values (3°, 3.7°, 4.3°, 5.0°, 5.7°, 6.4°). This resulted in 96 trials: 6 sizes × 2 velocities × 2 eyes × 4 trials per condition. To eliminate a potential “edge effect,” the edges of the stimuli were smoothed over 0.8° with a Gaussian function (see Supplementary Files S1S5 for stimulus examples). 
As before, the participants characterized both the velocity dominance and its spatial extent on every trial. There were four dominance outcomes: stationary dominance (“down” button), dynamic dominance (“up”), piecemeal rivalry (“left”), or transparency (“right”). 
Results
Stationary dominance (either whole-field or just the center) resulted more frequently from larger competing stimuli (b = 0.81, z = 10.44, p < 0.001; BF > 1026; Figure 3A). Smaller stimuli were more likely to lead to piecemeal rivalry (average effect: b = −0.86, z = −10.49, p < 0.001; BF > 1027). Dynamic dominance or transparency were rare: 6% and <1% of all trials, respectively. 
Figure 3
 
Experiment 2: The effect of stimulus size on motion dominance. (A) Proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent dominance responses as a function of stimulus size. The larger stimuli were more likely to produce stationary dominance. (B) The incidence of center-surround segmentation (orange curve) and the size of the stationary center (blue curve) as a function of stimulus size. The majority of the stationary-dominant trials had center-surround pattern. The size of the center was a constant proportion of the stimulus size. Dots are group averages; error bars reflect standard errors. See Supplementary Figure S4 for individual data (N = 5).
Figure 3
 
Experiment 2: The effect of stimulus size on motion dominance. (A) Proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent dominance responses as a function of stimulus size. The larger stimuli were more likely to produce stationary dominance. (B) The incidence of center-surround segmentation (orange curve) and the size of the stationary center (blue curve) as a function of stimulus size. The majority of the stationary-dominant trials had center-surround pattern. The size of the center was a constant proportion of the stimulus size. Dots are group averages; error bars reflect standard errors. See Supplementary Figure S4 for individual data (N = 5).
There was a reliable stationary-center/dynamic-surround spatial pattern in trials with stationary dominance (Figure 3B, orange curve). On trials in which the participants indicated stationary dominance (whole-field or center), the percept had a stationary center and fast surround in >90% of cases for all but the smallest stimulus size. There was no linear effect of stimulus size on the center-surround incidence (b = 0.23, z = 1.08, p = 0.282; BF = 0.18). Importantly, the stationary center scaled with the stimulus size: The size of the center remained constant at ∼72% of the stimulus diameter (SD = 1%, b = –0.01, t = –0.73, p = 0.264; BF = 0.15; Figure 3B, blue curve). 
Discussion
Manipulating the velocity bandwidth (Experiments 2 and onward) instead of directional velocity (Experiment 1) produced the same slow-center/fast-surround dominance pattern. Increasing the size of the interocular stimuli increased the proportion of trials with stationary center dominance (i.e., scale-variant). On the other hand, the stationary center scaled to the size of the two competitors (i.e., scale-invariant). In other words, although a decrease in size led to a decreased probability of the center-surround segmentation, the size of the central stationary dominance was independent of stimulus size. As stimulus edges were blurred to reduce potential “edge effect,” it is unlikely that the segmentation can be explained purely in terms of increased salience of dynamic stimuli along the edges. Because eccentricity alone cannot fully account for the center-surround segmentation, an explanation of the phenomenon should address both the scale-variant and scale-invariant properties. These explanations are considered in the General discussion
Note that the incidence of “piecemeal” rivalry declined with the stimulus size, which is precisely the opposite of the previously observed pattern (Blake et al., 1992). Typically, “piecemeal rivalry” refers to instances when parts of both monocular stimuli form the current percept. However, we distinguished between center-surround and piecemeal rivalry as opposed to clumping them together. In the current study, the “piecemeal” rivalry specifically referred to instances when the percept could not be characterized in terms of a center-surround spatial pattern. This may have led to an apparent discrepancy in our findings as the incidence of center-surround spatial segmentation rose as a function of stimulus size. 
Experiment 3
The focus of Experiments 3 and 4 was to investigate the temporal dimension of the stationary center/fast surround percept. In Experiment 3, we sought to establish the causality of the spatial segmentation by having either stationary or dynamic stimulus presented at various intervals following the onset of the other. Note that this paradigm is similar to flash suppression (Wolfe, 1984) and onset rivalry (Carter & Cavanagh, 2007; Ding, Naber, Gayet, Van der Stigchel, & Paffen, 2018; Stanley et al., 2011). The delays ranged from 0 (simultaneous onset) to 1 s, in 200-ms increments, making the trial duration vary from 1 to 2 s. The delayed stimulus was always presented for 1 s. This resulted in two nonsynchronous conditions: in the stationary delay condition, the stationary stimulus was introduced after the dynamic one and vice versa in the dynamic delay condition. One possibility is that the effect occurs exclusively with simultaneous offset; another is that it is triggered by the onset of the dynamic stimulus; yet another is that it is triggered by the onset of the stationary stimulus. 
Methods
The participants and methods were also the same as in Experiment 2 except that either the stationary or dynamic stimulus was introduced after its competitor. Stimulus asynchrony varied from 0 to 1,000 ms in 200-ms increments. Counterbalanced for each velocity and eye, this resulted in 6 asynchrony values × 2 asynchrony types (stationary or dynamic) × 2 eyes × 4 trials = 96 trials per run. 
Due to large individual differences (Figure 4), the GLM analysis was carried out with the consideration of the interindividual variation in the effect of interest, i.e., stimulus asynchrony/delay. To this end, the individual participants' slopes were included as the random factor (delay | participant) in the mixed models. Note that this is the more conservative measure. The effects were considered significant if they persisted in face of the interindividual variation. 
Figure 4
 
Experiment 3: The effect of stationary and dynamic stimulus delay on dominance. The leftmost graph is the average. (A) The proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent responses as a function of stationary stimulus delay. Varying the stationary stimulus asynchrony did not change the incidence of stationary dominance for the following second. (B) Proportion of different response types as a function of dynamic stimulus delay following stationary stimulus. Introducing dynamic stimulus >600 ms after stationary stimulus largely abolished stationary dominance for most participants in favor of piecemeal rivalry. Dots are group averages; error bars reflect standard errors.
Figure 4
 
Experiment 3: The effect of stationary and dynamic stimulus delay on dominance. The leftmost graph is the average. (A) The proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent responses as a function of stationary stimulus delay. Varying the stationary stimulus asynchrony did not change the incidence of stationary dominance for the following second. (B) Proportion of different response types as a function of dynamic stimulus delay following stationary stimulus. Introducing dynamic stimulus >600 ms after stationary stimulus largely abolished stationary dominance for most participants in favor of piecemeal rivalry. Dots are group averages; error bars reflect standard errors.
Results
As there was a large interparticipant variability, individual data are presented (Figure 4). 
We first examined the trials when the stationary stimulus was introduced after the onset of the dynamic stimulus (Figure 4A). Increasing onset asynchrony had no consistent impact on either whole-field or center-surround stationary dominance (b = 1.17, z = 0.82, p = 0.413; BF = 3.50 and b = −1.23, z = −1.27, p = 0.205; BF = 0.51, respectively). On average, the whole-field and center-surround stationary dominance accounted for 40% (SD = 9%) and 35% (SD = 10%) of these trials, respectively. 
On the other hand, introducing the dynamic stimulus after the stationary stimulus had an effect on the resulting dominance (Figure 4B). Specifically, the incidence of stationary center-surround dominance decreased with longer delay, and the proportion of piecemeal rivalry trials increased slightly (b = −2.95, z = −2.37, p = 0.018; BF > 10−9 and b = 1.59, z = 1.67, p = 0.095; BF > 10−6, respectively). No other effects were close to significance. 
The size of the stationary center on the stationary-dominant trials was 85% (SD = 2%) of the diameter for stationary delay and 78% (SD = 3%) for dynamic delay. Center size remained relatively constant across stationary and dynamic stimulus delays (BF = 1.09 and BF = 0.21, respectively). 
Discussion
If the dynamic stimulus appeared first, stationary dominance did not change with onset delays for most participants. On the other hand, displaying the dynamic stimulus more than 600 ms after the stationary stimulus onset generally disrupted stationary dominance in the center. 
However, none of these patterns held for all participants. Two out of five participants (S3 and S4) reported very few trials with whole-field stationary dominance. For participant S2, stationary dominance (whole-field or central) persisted even when the dynamic stimulus was introduced after the stationary one. Participant S6 reported most trials as whole-field stationary-dominant. Furthermore, the reported size of the stationary center differed from the previous experiment (85%/78% compared to 72% in Experiment 2). It is possible that this was due to the individual variability as well as stimulus manipulation. 
Reasons for such wide variability of results are unclear. It is possible that the asynchronous onsets of the two stimuli trigger more stochastic dominance patterns (Blake, 2001). If so, the differences seen here could be likened to the sustained rivalry paradigm rather than the onset rivalry as onset rivalry is known to be more consistent and less stochastic (Carter & Cavanagh, 2007; Stanley et al., 2011). 
Experiment 4
In Experiments 13, perceptual judgment about dominance had to be made at the end of each 1-s trial. Because the percept may have changed throughout this period, the participants were asked to characterize the latest percept. In Experiment 4, we sought to establish these perceptual changes by varying the duration of the trials. As before, the participants characterized the latest percept before the trial termination. Because the trials were of different durations, we could establish the timeline of dominance and spatial pattern. 
Methods
The methods were the same as in Experiment 3 with the following exceptions. Two participants (S2 and S3) who took part in previous experiments and six participants with no prior experience with this paradigm were recruited. Both the stationary and the dynamic stimulus were introduced simultaneously. The trial duration varied from 330 to 2,000 ms. The participants were instructed to report only the latest percept and disregard all the changes that occurred prior to that in their responses. 
Results
We first looked at the effect of trial duration. There was a significant drop in frequency of both whole-field and central stationary dominance for longer trials (b = −1.32, z = −3.76, p < 0.001; BF > 1011; both p values < 0.01 when evaluated separately; Figure 5A). Longer trials were instead associated with piecemeal rivalry or dynamic dominance (b = 1.56, z = −3.67, p < 0.001; BF > 107 and b = 0.73, z = 2.55, p = 0.011; BF = 5.73, respectively). Transparency was rare and did not change significantly with trial durations (M = 4%, SD = 2%; b = 2.04, z = 1.95, p = 0.051; BF = 0.10). This pattern of dominance was consistent across both naïve and experienced participants (see Supplemental Figure S5). 
Figure 5
 
Experiment 4: The effect of trial duration on motion dominance and center size. (A) Proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent responses as a function of trial duration. Trials up to 1 s in duration were more likely to result in stationary dominance at the end of the trial. (B) The proportion of center/surround trials among stationary-dominant trials (in orange) and the size of the stationary center (in blue) as a function of trial duration. The stationary stimulus dominated in the center but not the periphery on a majority of trials. As the trials increased in duration, the size of the stationary center decreased slightly. The size of the center is expressed as a proportion relative to the stimulus diameter. Dots are group averages; error bars reflect standard errors. See Supplemental Figure S5 for individual data (N = 8).
Figure 5
 
Experiment 4: The effect of trial duration on motion dominance and center size. (A) Proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent responses as a function of trial duration. Trials up to 1 s in duration were more likely to result in stationary dominance at the end of the trial. (B) The proportion of center/surround trials among stationary-dominant trials (in orange) and the size of the stationary center (in blue) as a function of trial duration. The stationary stimulus dominated in the center but not the periphery on a majority of trials. As the trials increased in duration, the size of the stationary center decreased slightly. The size of the center is expressed as a proportion relative to the stimulus diameter. Dots are group averages; error bars reflect standard errors. See Supplemental Figure S5 for individual data (N = 8).
We then analyzed the center-surround incidence and the center size. Only trials of 1 s and shorter (0.33–1 s) were considered for this analysis as these were the trials with stationary dominance occurring on more than 50% of the trials (see Figure 5A). A majority of stationary-dominant trials had a center-surround pattern across trial durations (M = 89%, SD = 7%; b = 0.66, z = 1.47, p = 0.141; BF = 0.33; Figure 5B, orange curve) with the center shrinking slightly throughout the first second (b = −0.08, t = −2.94, p = 0.003; BF = 12.30; Figure 5B, blue curve). 
Discussion
In Experiment 4, we found a consistent central stationary dominance bias in trials up to 1 s in duration, which was abolished thereafter, and the stochastic rivalry processes took over. Notably, the dynamic stimuli never fully overtook the stationary ones in dominance even in longer trials; instead, many piecemeal rivalry trials were observed (see Figure 5A). 
The size of the stationary center also decreased slightly but consistently across time. This points to the dynamic nature of the phenomenon, which is not statically linked to eccentricity. Instead, the central dominance gradually receded until it disappeared around the 1-s mark. 
Experiment 5
Experiments 14 showed that the stationary onset dominance was confined to the center of interocular stimuli. Experiment 2 demonstrated that the effect depends at least partially on eccentricity. We reasoned that the prioritization of the stationary signal in the center of the stimulus could be explained in at least three ways. First, it is possible that the stationary signal is overall strong but weaker along the border when competing with fast stimulus. Conversely, it is possible that the stationary signal is spatially homogenous while the fast signal is weaker in the center. In other words, in these two cases, the spatial inhomogeneity may be driven by either the stationary or fast signal. Finally, it is possible that both stationary and fast stimuli are spatially homogenous, and the center-surround split is purely driven by the interocular competition (e.g., interocular pooling or normalization; see Li, Carrasco, & Heeger, 2015; Ling & Blake, 2012). 
To probe these possibilities, in Experiment 5, we used an interocular suppression paradigm. Motion clouds moving at different velocities (1°/s to 8°/s) acted as interocular masks in one eye while either a stationary or fast (0°/s or 16°/s) target was presented in the other eye. To explore the spatial effects, we varied the target eccentricity. As the targets were matched to the masks in terms of surface features, we expected there to be a feature-selective suppression for velocity (Han & Alais, 2018): Higher mask speeds should be more effective at suppressing fast targets; slower masks should lead to higher detection thresholds for stationary targets. The expected effects of target eccentricity were, thus, expected to emerge above and beyond feature-selective suppression. They would manifest as differences in suppression effectiveness at different eccentricities as a function of both target and mask velocities. 
Methods
Eight participants, including the first author, took part in the experiment. One more participant took part in previous experiments (14). All except the first author were naïve as to the purpose of the experiment. One participant (#1) was excluded due to a technical issue; another (#3) was excluded because of failure to converge. The analyses were run on the remaining six participants' data. 
The speeds of motion cloud masks were 1°/s, 2°/s, 4°/s, and 8°/s in velocity bandwidth (Figure 6). Other mask parameters (size, edge fade) were the same as those used in Experiments 14
Figure 6
 
An example trial in Experiment 5. Each trial started with a simultaneous onset of a mask to the dominant eye and a target to the nondominant eye. Mask speed bandwidth was randomly assigned to 1°/s, 2°/s, 4°/s, or 8°/s, and the target was either stationary or fast (0°/s or 16°/s, respectively). In this example trial, the target was “near” at 0.9° off-center; a “far” target would be presented closer to the edge at 2.7° off-center.
Figure 6
 
An example trial in Experiment 5. Each trial started with a simultaneous onset of a mask to the dominant eye and a target to the nondominant eye. Mask speed bandwidth was randomly assigned to 1°/s, 2°/s, 4°/s, or 8°/s, and the target was either stationary or fast (0°/s or 16°/s, respectively). In this example trial, the target was “near” at 0.9° off-center; a “far” target would be presented closer to the edge at 2.7° off-center.
The targets were motion cloud patches 0.5° in size. They were either stationary (0°/s) or fast (16°/s), corresponding to the stationary and dynamic competing patterns used in previous experiments. The target location varied in eccentricity (0.9° and 2.7°) and side (left/right). The target onset coincided with the onset of the mask to replicate the conditions for the onset rivalry in earlier experiments. The contrast of the target was 100% at the onset and faded out linearly for 400 ms. A trial lasted 600 ms. 
On each trial, the participants were to make a forced judgment of whether the target appeared to the left or right of the center. The contrast detection threshold for target Gabors was determined using a one-up, one-down staircase method with 10 reversals as the termination criterion (Kingdom & Prins, 2010). The up-to-down step size ratio was 0.74, targeting 83% accuracy (García-Pérez, 1998). 
Each participant's eye dominance was determined by measuring the contrast detection threshold for both eyes under the conditions approximating the main experiment (Ding et al., 2018). The threshold experiment used an “intermediate” velocity bandwidth of 4°/s cloud mask with the “central” target, presented 0.9° left or right off-center. The target eye with the lower threshold was deemed to be the dominant eye. In the main experiment, the mask was always presented to that eye. 
The following full-interaction formula was used for both GLM and BF analyses with participant's intercept as a random factor:  
\(\def\upalpha{\unicode[Times]{x3B1}}\)\(\def\upbeta{\unicode[Times]{x3B2}}\)\(\def\upgamma{\unicode[Times]{x3B3}}\)\(\def\updelta{\unicode[Times]{x3B4}}\)\(\def\upvarepsilon{\unicode[Times]{x3B5}}\)\(\def\upzeta{\unicode[Times]{x3B6}}\)\(\def\upeta{\unicode[Times]{x3B7}}\)\(\def\uptheta{\unicode[Times]{x3B8}}\)\(\def\upiota{\unicode[Times]{x3B9}}\)\(\def\upkappa{\unicode[Times]{x3BA}}\)\(\def\uplambda{\unicode[Times]{x3BB}}\)\(\def\upmu{\unicode[Times]{x3BC}}\)\(\def\upnu{\unicode[Times]{x3BD}}\)\(\def\upxi{\unicode[Times]{x3BE}}\)\(\def\upomicron{\unicode[Times]{x3BF}}\)\(\def\uppi{\unicode[Times]{x3C0}}\)\(\def\uprho{\unicode[Times]{x3C1}}\)\(\def\upsigma{\unicode[Times]{x3C3}}\)\(\def\uptau{\unicode[Times]{x3C4}}\)\(\def\upupsilon{\unicode[Times]{x3C5}}\)\(\def\upphi{\unicode[Times]{x3C6}}\)\(\def\upchi{\unicode[Times]{x3C7}}\)\(\def\uppsy{\unicode[Times]{x3C8}}\)\(\def\upomega{\unicode[Times]{x3C9}}\)\(\def\bialpha{\boldsymbol{\alpha}}\)\(\def\bibeta{\boldsymbol{\beta}}\)\(\def\bigamma{\boldsymbol{\gamma}}\)\(\def\bidelta{\boldsymbol{\delta}}\)\(\def\bivarepsilon{\boldsymbol{\varepsilon}}\)\(\def\bizeta{\boldsymbol{\zeta}}\)\(\def\bieta{\boldsymbol{\eta}}\)\(\def\bitheta{\boldsymbol{\theta}}\)\(\def\biiota{\boldsymbol{\iota}}\)\(\def\bikappa{\boldsymbol{\kappa}}\)\(\def\bilambda{\boldsymbol{\lambda}}\)\(\def\bimu{\boldsymbol{\mu}}\)\(\def\binu{\boldsymbol{\nu}}\)\(\def\bixi{\boldsymbol{\xi}}\)\(\def\biomicron{\boldsymbol{\micron}}\)\(\def\bipi{\boldsymbol{\pi}}\)\(\def\birho{\boldsymbol{\rho}}\)\(\def\bisigma{\boldsymbol{\sigma}}\)\(\def\bitau{\boldsymbol{\tau}}\)\(\def\biupsilon{\boldsymbol{\upsilon}}\)\(\def\biphi{\boldsymbol{\phi}}\)\(\def\bichi{\boldsymbol{\chi}}\)\(\def\bipsy{\boldsymbol{\psy}}\)\(\def\biomega{\boldsymbol{\omega}}\)\(\def\bupalpha{\unicode[Times]{x1D6C2}}\)\(\def\bupbeta{\unicode[Times]{x1D6C3}}\)\(\def\bupgamma{\unicode[Times]{x1D6C4}}\)\(\def\bupdelta{\unicode[Times]{x1D6C5}}\)\(\def\bupepsilon{\unicode[Times]{x1D6C6}}\)\(\def\bupvarepsilon{\unicode[Times]{x1D6DC}}\)\(\def\bupzeta{\unicode[Times]{x1D6C7}}\)\(\def\bupeta{\unicode[Times]{x1D6C8}}\)\(\def\buptheta{\unicode[Times]{x1D6C9}}\)\(\def\bupiota{\unicode[Times]{x1D6CA}}\)\(\def\bupkappa{\unicode[Times]{x1D6CB}}\)\(\def\buplambda{\unicode[Times]{x1D6CC}}\)\(\def\bupmu{\unicode[Times]{x1D6CD}}\)\(\def\bupnu{\unicode[Times]{x1D6CE}}\)\(\def\bupxi{\unicode[Times]{x1D6CF}}\)\(\def\bupomicron{\unicode[Times]{x1D6D0}}\)\(\def\buppi{\unicode[Times]{x1D6D1}}\)\(\def\buprho{\unicode[Times]{x1D6D2}}\)\(\def\bupsigma{\unicode[Times]{x1D6D4}}\)\(\def\buptau{\unicode[Times]{x1D6D5}}\)\(\def\bupupsilon{\unicode[Times]{x1D6D6}}\)\(\def\bupphi{\unicode[Times]{x1D6D7}}\)\(\def\bupchi{\unicode[Times]{x1D6D8}}\)\(\def\buppsy{\unicode[Times]{x1D6D9}}\)\(\def\bupomega{\unicode[Times]{x1D6DA}}\)\(\def\bupvartheta{\unicode[Times]{x1D6DD}}\)\(\def\bGamma{\bf{\Gamma}}\)\(\def\bDelta{\bf{\Delta}}\)\(\def\bTheta{\bf{\Theta}}\)\(\def\bLambda{\bf{\Lambda}}\)\(\def\bXi{\bf{\Xi}}\)\(\def\bPi{\bf{\Pi}}\)\(\def\bSigma{\bf{\Sigma}}\)\(\def\bUpsilon{\bf{\Upsilon}}\)\(\def\bPhi{\bf{\Phi}}\)\(\def\bPsi{\bf{\Psi}}\)\(\def\bOmega{\bf{\Omega}}\)\(\def\iGamma{\unicode[Times]{x1D6E4}}\)\(\def\iDelta{\unicode[Times]{x1D6E5}}\)\(\def\iTheta{\unicode[Times]{x1D6E9}}\)\(\def\iLambda{\unicode[Times]{x1D6EC}}\)\(\def\iXi{\unicode[Times]{x1D6EF}}\)\(\def\iPi{\unicode[Times]{x1D6F1}}\)\(\def\iSigma{\unicode[Times]{x1D6F4}}\)\(\def\iUpsilon{\unicode[Times]{x1D6F6}}\)\(\def\iPhi{\unicode[Times]{x1D6F7}}\)\(\def\iPsi{\unicode[Times]{x1D6F9}}\)\(\def\iOmega{\unicode[Times]{x1D6FA}}\)\(\def\biGamma{\unicode[Times]{x1D71E}}\)\(\def\biDelta{\unicode[Times]{x1D71F}}\)\(\def\biTheta{\unicode[Times]{x1D723}}\)\(\def\biLambda{\unicode[Times]{x1D726}}\)\(\def\biXi{\unicode[Times]{x1D729}}\)\(\def\biPi{\unicode[Times]{x1D72B}}\)\(\def\biSigma{\unicode[Times]{x1D72E}}\)\(\def\biUpsilon{\unicode[Times]{x1D730}}\)\(\def\biPhi{\unicode[Times]{x1D731}}\)\(\def\biPsi{\unicode[Times]{x1D733}}\)\(\def\biOmega{\unicode[Times]{x1D734}}\)\begin{equation}threshold\sim maskSpeed \times targetVelocity \times targetEccentricity + (1|observer)\end{equation}
 
The BF values were computed for each regression estimate by comparing models with and without a given factor (option “whichModels” in BayesFactor package function “generalTestBF” set to “top”). The resulting BF values were inverted, such that the nominator model was the complete model while the denominator model excluded a given factor of interest. 
We also ran complete models that included the participant's dominant eye and the starting value of the staircase as nuisance factors. As they did not change the effects of interest and for simplicity, we only report the non-nuisance model results. 
Results
The results for the main effects and interactions from GLM and Bayes factor are shown in Table 2 and Figure 7 (individual data in Supplementary Figure S6). The faster masks were less effective overall, confirming previous reports (effect #1; Ananyev, Penney, & Hsieh, 2017; Han, Blake, & Alais, 2018). Moving targets had lower contrast thresholds, i.e., were easier to detect, than the stationary target (effect #2). Peripheral targets were slightly harder to spot than central targets overall (effect #3). 
Table 2
 
Regression and Bayes factor results of the mask speed, target speed, and target eccentricity. Notes: P and BF values in bold are significant or approaching significance. Higher BF values indicate support of the model that includes the corresponding variable (nominator model) compared to the model that excludes it (denominator model).
Table 2
 
Regression and Bayes factor results of the mask speed, target speed, and target eccentricity. Notes: P and BF values in bold are significant or approaching significance. Higher BF values indicate support of the model that includes the corresponding variable (nominator model) compared to the model that excludes it (denominator model).
Figure 7
 
Contrast detection threshold as a function of mask speed, target speed, and target eccentricity. Stationary targets were more effectively suppressed by slower masks and vice versa for fast targets.
Figure 7
 
Contrast detection threshold as a function of mask speed, target speed, and target eccentricity. Stationary targets were more effectively suppressed by slower masks and vice versa for fast targets.
Of special interest were the interactions (effects #4–7). Dynamic (fast) targets were markedly harder to detect with faster masks (effect #4). Conversely, stationary targets were better suppressed by slower masks. This can be seen in Figure 7 as an increase in contrast threshold for dynamic targets and decrease for stationary targets with an increase in mask speed (left to right). This indicated a feature-selective suppression effect consistent with previous reports (Han & Alais, 2018; Moors, Wagemans, & de-Wit, 2014; Yang & Blake, 2012). 
Importantly, there were significant interactions with target eccentricity (effects #5–7). Specifically, dynamic targets were harder to detect in periphery compared to central presentation, and stationary targets were harder to detect centrally (effect #6). This resulted in target eccentricity having generally positive and negative slopes for dynamic and stationary targets, respectively. The slopes are seen to intersect at intermediate mask velocity (see 4°/s mask in Figure 7). 
The three-way interaction between target and mask velocities and target eccentricity was also significant (effect #7). This was above and beyond other effects in the model. The interaction can be seen as the decrease of eccentricity effects for fast target/fast mask, i.e., the “shallower” blue line slope for the fast 8°/s mask (Figure 7). 
The evidence for the analogous target eccentricity × mask speed interaction for the stationary target was ambiguous. The increase in eccentricity slope for the stationary target with higher mask speed reached significance in GLM but not Bayesian analysis (effect #5). The slope for the stationary target is seen as more positive at higher mask speeds (rightmost plot in Figure 7). The discrepancy between the model outcomes was likely due to the GLM being run on binary variables (effect #5 is for stationary target as the intercept target speed is zero), whereas the statistical package used for Bayesian analyses centered the variables (effect #5 across target speeds as zero is the average). This was confirmed by running both GLM and Bayesian analyses on centered variables: Although effect #5 became less significant in GLM analyses (t = 2.03, P = 0.042), the BF remained largely unchanged (BF = 0.65). We, thus, interpreted the evidence of the eccentricity–mask speed interaction as positive. 
Discussion
As expected, the fast target was more effectively suppressed by a fast mask and vice versa for the stationary target. This is a hallmark of speed-selective suppression (Han & Alais, 2018; Moors et al., 2014; Yang & Blake, 2012). 
The primary objective of this experiment, however, was to establish whether slow and/or fast motion fields were spatially homogenous in suppression strength. Indeed, we found that the target eccentricity effects were more pronounced for the slower masks. Conversely, the target eccentricity slopes were shallower for faster masks (compare slopes for 1°/s and 8°/s masks in Figure 7). This suggests that the slow-center/fast-surround segmentation observed in binocular rivalry (Experiments 14) could potentially be explained by the spatial inhomogeneity of the slow but not fast motion channel. This possibility is further supported by the fact that the stationary target slope changed signs at fastest mask speed, and the fast target slope remained positive. 
General discussion
When a stationary or slow cloud pattern engaged in interocular competition with a fast motion cloud, the slower stimulus dominated following onset, but only centrally (Experiment 1). Furthermore, two opposing motions often cancelled out, leading to a whole-field stationary percept. Both the directional velocity (Experiment 1) and velocity bandwidth manipulation (Experiment 2 and onward) consistently produced a slow-center/fast-surround dominance pattern. The center-surround spatial segmentation was less common for smaller competing stimuli, but the stationary center size remained a constant proportion of the stimulus diameter regardless of stimulus size (Experiment 2). Although the central stationary onset bias was disrupted by the dynamic stimulus asynchrony, its occurrence was specifically linked to the onset of the stationary stimulus (Experiment 3) and lasted for approximately 1 s (Experiment 4). In Experiment 5, we used a target-detection paradigm to demonstrate an analogous effect in the context of interocular suppression. Consistent with the binocular rivalry findings, the eccentricity of a target interacted with speeds of both the target and the mask in interocular suppression. 
We discuss two novel effects of particular interest: (a) the general bias toward stationary stimuli, resulting in motion cancelling if the opposing velocities were low, and (b) the link between the stationary interocular bias and eccentricity. In Experiment 1, if the competing directional velocities were sufficiently low, they often led to a nondirection (stationary) percept. This is similar to an effect reported on binocular viewing of opposing spiral motions cancelling each other out (de Weert & Wade, 1984). The authors of that study concluded that the extent of the cancelling depended on a so-called “fusion limit.” The motion stimuli may be reconciled into a nonmoving percept if the distance between stimuli do not exceed some threshold, similar to how binocular stimuli can only be fused if they are close to each other in retinal space. The motion clouds used here may have facilitated this reconciliation due to the unstructured configuration of motion clouds minimizing form information (Alais & Parker, 2006, 2012). Due to our use of directional motion, if the velocities were high enough, they were less likely to be reconciled into a stationary percept. This may have been due to the exceeding of the “fusion limit” (reconciliation threshold). Support for this interpretation comes from models of binocular rivalry that incorporate fusion by including two populations of neurons: binocular summation neurons that normalize input across the two eyes and inhibitory “opponency” neurons that produce rivalry (Said & Heeger, 2013; Wilson, 2017). 
A compelling explanation for the second effect, the center-surround motion-based segmentation, should address why the stationary bias is specific to the center of the stimulus space. Experiment 3 hinted at the causal contribution of the stationary stimulus in particular: The central stationary bias was present even if the stationary stimulus was introduced after the dynamic stimulus. Experiment 5 confirmed that the slower motion channel was nonhomogenous across the visual field. This conclusion was made based on the following two findings. First, target eccentricity mattered less for faster masks (shallower slopes). Second, the eccentricity effect reversed signs with higher mask speed for the stationary but not the fast target. These effects imply the causal role of the slow-motion channel in the observed spatial inhomogeneity (the center-surround pattern) in dominance. 
It is possible to account for some of the obtained data with the known variability of contrast sensitivity to temporal frequencies. The lower the spatial frequency, the higher the contrast sensitivity to high speeds (Kelly, 1979). As the perceived spatial frequency is lower in the periphery, this could potentially explain the dominance of high velocities along the edges of the stimuli. This explanation, however, was only partly supported with our data. First, per Kelly's data, the intermediate velocities should have been more dominant than we observed. In fact, intermediate velocities rarely dominated in Experiment 1 (<20% of trials). Second, although decreasing the size of the stimulus did decrease the number of center-surround trials, the stationary center dominance remained constant in size (Experiment 2), suggesting that the center stationary bias is not completely dependent on eccentricity. 
Two classes of models may help explain the observed interocular stationary bias and its connection with eccentricity: two-channel models and divisive normalization. 
Two-channel models are frequently used to account for different outcomes based on stimulus features, such as motion speed (Bullier & Nowak, 1995; Himmelberg & Wade, 2019; Stone, Dreher, & Leventhal, 1979). For example, a two-channel model was used to characterize the differences in fMRI-measured responses to changes in luminance (Horiguchi, Nakadomari, Misaki, & Wandell, 2009). Horiguchi et al. (2009) used two temporal channels, transient and sustained, to model the differences in BOLD response at 10° and 40° eccentricities. In our case, the two-channel model could help explain the link between eccentricity and velocity. This is because it posits that some features, such as eccentricity, color, and spatial and temporal frequencies, are processed by two different channels and are, thus, bound within those channels, such as with parvocellular and magnocellular pathways. The two pathways originate in the retina, go through lateral geniculate nucleus (LGN) and extrastriate areas, and subsequently predominantly fuse into ventral and dorsal pathways, respectively (Breitmeyer, 2014; Livingstone & Hubel, 1987; Nassi & Callaway, 2009; Tapia & Breitmeyer, 2011). Parvocellular pathway is known to be biased toward higher spatial and lower temporal frequencies and is color-opponent; magnocellular pathway, on the other hand, is associated with lower spatial and higher temporal frequencies and is color-blind. Livingstone and Hubel (1988, p. 240) proposed that the more primitive “magno system is not capable of sustained scrutiny” while “the parvo system seems to be important for analyzing the scene in much greater and more leisurely detail.” This helps explain eccentricity-related effects obtained here as the ratio of parvocellular to magnocellular neurons in LGN is known to decline with eccentricity (from ∼35 in the center to ∼15 five degrees of visual angle off center), such that the central processing is predominantly parvo-biased (Azzopardi, Jones, & Cowey, 1999). 
However, the fact that the size of the stationary center scaled to the overall stimulus size in Experiment 2 points to the insufficiency of eccentricity-based explanation. Although the stationary-center/fast-surround pattern was less frequent for smaller stimuli, it was scale-invariant whenever it occurred. The dominance did not seem to be determined, for instance, by the higher foveal acuity as such, as the stimulus, being approximately 6° in diameter, was foveally projected. Nor was it dictated by the interactions of motion signal with hard edges as edges were blurred in Experiments 24
Divisive normalization has been recently used to model interocular and binocular effects (Li et al., 2015; Quaia, Optican, & Cumming, 2017, 2018; Said & Heeger, 2013). Divisive normalization is the division (normalization) of the neural signal by a larger signal pool (Carandini & Heeger, 2012). It can be thought of as a way the brain fine-tunes the signal by considering its spatial and/or temporal context. Divisive normalization has been used to account for a wide variety of phenomena, including light adaptation (Carandini & Heeger, 2012), cross-orientation masking (Foley, 1994), surround masking (Petrov, Carandini, & McKee, 2005), and neural adaptation (Zhou, Benson, Kay, & Winawer, 2018). 
Divisive normalization explains the effectiveness of the mask based on its featural similarity to the target (Carandini & Heeger, 2012; Foley, 1994; Li et al., 2015; Petrov et al., 2005; Quaia et al., 2017; Said & Heeger, 2013). In our case, the divisive normalization model could account for the feature-selective suppression observed in Experiment 5: The slower masks more effectively suppressed the stationary target, and the faster masks were more effective at suppressing the fast target (mask speed × target speed interaction). Additionally, it may account for the observed size-invariance, i.e., the constant proportion of the stationary center in relation to the fast surround (and stimulus size). This phenomenon may occur due to surround inhibition: The inhibitory signal at the edges includes empty space outside of the bounds of the stimulus. The link between surround suppression and binocular rivalry has been recently computationally described (Li et al., 2015). Finally, normalization is a component in binocular rivalry models that account for fusion behavior (Said & Heeger, 2013; Wilson, 2017), which helps explain the motion canceling observed in Experiment 1 at low velocities (see above). 
To sum up the above speculations, the two-channel models in conjunction with divisive normalization may be able to account for the observed central dominance of slow or stationary stimuli. Future computational work is needed to establish the causes and neural underpinnings of these phenomena. 
Further work on motion-based spatial inhomogeneity should address several limitations of the current study. First, although the interocular suppression paradigm (Experiment 5) suggests that attention was not an important factor in binocular rivalry experiments (14), the role of attention has not been completely ruled out. Onset rivalry was shown to be robust against attentional influences in some cases (Chong & Blake, 2006; Dogge et al., 2018) while being a factor in others (Chong & Blake, 2006; Mitchell, Stoner, & Reynolds, 2004). Second, it is important to establish how much of the observed phenomenon is due to the interocular mechanisms. This can be done by investigating monocular and binocular presentations. This may prove important as speed can be used to segregate motion into different depth planes (Watamaniuk & Duchon, 1992). Third, we attribute the novelty of the eccentricity findings to the fact that we used a class of stimuli that minimized form information (Alais & Parker, 2006, 2012). However, an important future direction is to determine exactly which parameters lead to the departure from the rivalry observed with the popular random dot kinematograms (Blake et al., 1985; Marshak & Sekuler, 1979; van de Grind et al., 2001; Wiesenfelder & Blake, 1990). Computational work suggests that we should expect differences in dominance patterns if stimulus parameters are varied (Medathati et al., 2017). The motion cloud configuration used here corresponds to an intermediate motion variance between pure gratings (single spatial frequency) and random dot patterns (broadband stimulus). Medathati et al.'s (2017) work, thus, suggests that spatial frequency range is important, and this should be addressed in the future. 
One observed effect in Experiment 5 might be of concern for the interpretation of interocular suppression studies. Depending on the speed of the mask, the central targets were either more (with slow masks) or less (with fast masks) effectively suppressed than peripheral targets. This goes against a tacitly and frequently made assumption that the mask characteristics should not interact with different target features (here, eccentricity). For example, it is common to claim that certain target features are “prioritized for conscious access,” i.e., easier/faster to detect (e.g., Gayet, Paffen, & Van der Stigchel, 2013; Gobbini et al., 2013; Hedger, Gray, Garner, & Adams, 2016). Therefore, if the target eccentricity was the feature of interest, we may conclude that the central targets are prioritized for conscious access—but only if the mask used was slow. However, this conclusion would flip with faster masks, in which case the peripheral targets are “prioritized.” Naturally, such basic characteristics as target eccentricity and speed are rarely a concern of such studies. However, what is questioned here is the logic of the “prioritization for conscious access” argument itself. Target detection does not take place in a vacuum; instead, the signal is a function of the noise that masks it. This is vividly demonstrated by the feature-selective suppression in Experiment 5. This adds to related concerns with regards to interocular suppression (Kerr, Hesselmann, Räling, Wartenburger, & Sterzer, 2017; Moors, Boelens, van Overwalle, & Wagemans, 2016; Moors, Hesselmann, Wagemans, & van Ee, 2017; Stein et al., 2011). 
To conclude, the present findings demonstrate the importance of the interaction of two factors in binocular rivalry of moving and stationary stimuli: the relative temporal onset and the relative spatial position of the two stimuli. We showed that, when a stationary stimulus and sufficiently fast motion field were engaged in interocular competition, this resulted in a spontaneous and largely scale-invariant segmentation of the stimulus into a stationary center and dynamic surround lasting up to 1 s following the onset. These findings have wide-ranging implications for interocular suppression studies that often rely on a dynamic mask to suppress a stationary stimulus interocularly as well as for binocular rivalry in general, where the spatial pattern of dominance needs to be considered for a more complete understanding of interocular competition and the cortical mechanisms involved. 
Acknowledgments
We thank the Yushan Young Scholar Program (NTU-108V0202) for funding. 
Commercial relationships: none. 
Corresponding authors: Egor Ananyev; Po-Jang Hsieh. 
Address: Nanyang Technological University, Department of Psychology, Singapore; National Taiwan University, Department of Psychology, Taipei, Taiwan. 
References
Alais, D., & Parker, A. (2006). Independent binocular rivalry processes for motion and form. Neuron, 52 (5), 911–920, https://doi.org/10.1016/j.neuron.2006.10.027.
Alais, D., & Parker, A. (2012). Binocular rivalry produced by temporal frequency differences. Frontiers in Human Neuroscience, 6: 227, https://doi.org/10.3389/fnhum.2012.00227.
Ananyev, E., Penney, T. B., & Hsieh, P.-J. (2017). Separate requirements for detection and perceptual stability of motion in interocular suppression. Scientific Reports, 7 (1): 7230, https://doi.org/10.1038/s41598-017-07805-5.
Apthorp, D. M. (2011). The role of motion streaks in human visual motion perception. Retrieved from https://ses.library.usyd.edu.au/handle/2123/7432.
Apthorp, D. M., & Alais, D. (2009). Tilt aftereffects and tilt illusions induced by fast translational motion: Evidence for motion streaks. Journal of Vision, 9 (1): 27, 1–11, https://doi.org/10.1167/9.1.27. [PubMed] [Article]
Azzopardi, P., Jones, K. E., & Cowey, A. (1999). Uneven mapping of magnocellular and parvocellular projections from the lateral geniculate nucleus to the striate cortex in the macaque monkey. Vision Research, 39 (13), 2179–2189, https://doi.org/10.1016/S0042-6989(98)00319-8.
Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language, 59 (4), 390–412, https://doi.org/10.1016/j.jml.2007.12.005.
Bates, D., Mächler, M., Bolker, B. M., & Walker, S. C. (2015). Fitting linear mixed effects models using lme4. Journal of Statistical Software, 67 (1), 1–48, https://doi.org/10.18637/jss.v067.i01.
Blake, R. (2001). A primer on binocular rivalry, including current controversies. Brain and Mind, 2 (1), 5–38.
Blake, R., & Fox, R. (1974). Binocular rivalry suppression: Insensitive to spatial frequency and orientation change. Vision Research, 14 (8), 687–692, https://doi.org/10.1016/0042-6989(74)90065-0.
Blake, R., & Logothetis, N. K. (2002). Visual competition. Nature Reviews Neuroscience, 3 (1), 13–21, https://doi.org/10.1038/nrn701.
Blake, R., O'Shea, R. P., & Mueller, T. J. (1992). Spatial zones of binocular rivalry in central and peripheral vision. Visual Neuroscience, 8 (5), 469–478, https://doi.org/10.1017/S0952523800004971.
Blake, R., Yu, K., Lokey, M., & Norman, H. (1998). Binocular rivalry and motion perception. Journal of Cognitive Neuroscience, 10 (1), 46–60, https://doi.org/10.1162/089892998563770.
Blake, R., Zimba, L., & Williams, D. (1985). Visual motion, binocular correspondence and binocular rivalry. Biological Cybernetics, 52 (6), 391–397, https://doi.org/10.1007/BF00449596.
Breese, B. B. (1899). On inhibition. The Psychological Review: Monograph Supplements, 3 (1), i–65, https://doi.org/10.1037/h0092990.
Breitmeyer, B. G. (2014). Contributions of magno- and parvocellular channels to conscious and non-conscious vision. Philosophical Transactions of the Royal Society B: Biological Sciences, 369 (1641): 20130213, https://doi.org/10.1098/rstb.2013.0213.
Bullier, J., & Nowak, L. G. (1995). Parallel versus serial processing: New vistas on the distributed organization of the visual system. Current Opinion in Neurobiology, 5 (4), 497–503, https://doi.org/10.1016/0959-4388(95)80011-5.
Carandini, M., & Heeger, D. J. (2012). Normalization as a canonical neural computation. Nature Reviews Neuroscience, 13 (1), 51–62, https://doi.org/10.1038/nrn3136.
Carlson, T. A., & He, S. (2000). Visible binocular beats from invisible monocular stimuli during binocular rivalry. Current Biology, 10 (17), 1055–1058, https://doi.org/10.1016/S0960-9822(00)00672-2.
Carter, O., & Cavanagh, P. (2007). Onset rivalry: Brief presentation isolates an early independent phase of perceptual competition. PLoS One, 2 (4): e343, https://doi.org/10.1371/journal.pone.0000343.
Chong, S. C., & Blake, R. (2006). Exogenous attention and endogenous attention influence initial dominance in binocular rivalry. Vision Research, 46 (11), 1794–1803.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple correlation/regression analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Lawrence Erlbaum Associates.
Cowey, A., & Rolls, E. T. (1974). Human cortical magnification factor and its relation to visual acuity. Experimental Brain Research, 21 (5), 447–454, https://doi.org/10.1007/BF00237163.
Crick, F. (1996, February 8). Visual perception: Rivalry and consciousness. Nature, 379 (6565), 485–486, https://doi.org/10.1038/379485a0.
Crick, F., & Koch, C. (1998). Consciousness and neuroscience. Cerebral Cortex (New York, N.Y.: 1991), 8 (2), 97–107.
De Bruyn, B. (1997). Blending transparent motion patterns in peripheral vision. Vision Research, 37 (5), 645–648, https://doi.org/10.1016/S0042-6989(96)00117-4.
de Weert, C. M. M., & Wade, N. J. (1984). Cancelling of pattern motion: Dichoptic and monoptic observations. Vision Research, 24 (8), 883–887, https://doi.org/10.1016/0042-6989(84)90162-7.
Ding, Y., Naber, M., Gayet, S., Van der Stigchel, S., & Paffen, C. L. E. (2018). Assessing the generalizability of eye dominance across binocular rivalry, onset rivalry, and continuous flash suppression. Journal of Vision, 18 (6): 6, 1–13, https://doi.org/10.1167/18.6.6. [PubMed] [Article]
Dogge, M., Gayet, S., Custers, R., & Aarts, H. (2018). The influence of action-effect anticipation on bistable perception: Differences between onset rivalry and ambiguous motion. Neuroscience of Consciousness, 2018 (1): niy004, https://doi.org/10.1093/nc/niy004.
Foley, J. M. (1994). Human luminance pattern-vision mechanisms: Masking experiments require a new model. Journal of the Optical Society of America.A, Optics, image science, and vision, 11 (6), 1710–1719, https://doi.org/10.1364/JOSAA.11.001710.
Fox, R., & Check, R. (1968). Detection of motion during binocular rivalry suppression. Journal of Experimental Psychology, 78 (3, Pt.1), 388–395, https://doi.org/10.1037/h0026440.
García-Pérez, M. A . (1998). Forced-choice staircases with fixed step sizes: Asymptotic and small-sample properties. Vision Research, 38 (12), 1861–1881.
Gayet, S., Paffen, C. L., & Van der Stigchel, S. (2013). Information matching the content of visual working memory is prioritized for conscious access. Psychological Science, 24 (12), 2472–2480, https://doi.org/10.1177/0956797613495882.
Giles, N., Lau, H., & Odegaard, B. (2016). What type of awareness does binocular rivalry assess? Trends in Cognitive Sciences, 20 (10), 719–720.
Gobbini, M. I., Gors, J. D., Halchenko, Y. O., Rogers, C., Guntupalli, J. S., Hughes, H., & Cipolli, C. (2013). Prioritized detection of personally familiar faces. PLoS One, 8 (6): e66620.
Han, S., & Alais, D. (2018). Strength of continuous flash suppression is optimal when target and masker modulation rates are matched. Journal of Vision, 18 (3): 3, 1–14, https://doi.org/10.1167/18.3.3. [PubMed] [Article]
Han, S., Blake, R., & Alais, D. (2018). Slow and steady, not fast and furious: Slow temporal modulation strengthens continuous flash suppression. Consciousness and Cognition, 58, 10–19, https://doi.org/10.1016/j.concog.2017.12.007.
Hedger, N., Gray, K. L., Garner, M., & Adams, W. J. (2016). Are visual threats prioritized without awareness? A critical review and meta-analysis involving 3 behavioral paradigms and 2696 observers. Psychological Bulletin, 142 (9), 934–968.
Himmelberg, M. M., & Wade, A. R. (2019). Eccentricity-dependent temporal contrast tuning in human visual cortex measured with fMRI. NeuroImage, 184, 462–474, https://doi.org/10.1016/j.neuroimage.2018.09.049.
Hong, S. W., & Blake, R. (2009). Interocular suppression differentially affects achromatic and chromatic mechanisms. Attention, Perception, & Psychophysics, 71 (2), 403–411, https://doi.org/10.3758/APP.71.2.403.
Horiguchi, H., Nakadomari, S., Misaki, M., & Wandell, B. A. (2009). Two temporal channels in human V1 identified using fMRI. NeuroImage, 47 (1), 273–280, https://doi.org/10.1016/j.neuroimage.2009.03.078.
Jeffreys, H. (1939). The theory of probability (1st ed.). Oxford: Oxford University Press.
Kelly, D. H. (1979). Motion and vision. II. Stabilized spatio-temporal threshold surface. JOSA, 69 (10), 1340–1349, https://doi.org/10.1364/JOSA.69.001340.
Kerr, J. A., Hesselmann, G., Räling, R., Wartenburger, I., & Sterzer, P. (2017). Choice of analysis pathway dramatically affects statistical outcomes in breaking continuous flash suppression. Scientific Reports, 7 (1): 3002, https://doi.org/10.1038/s41598-017-03396-3.
Kingdom, F. A. A., & Prins, N. (2010). Psychophysics: A practical introduction. London: Academic Press.
Kovács, I., Papathomas, T. V., Yang, M., & Fehér, Á. (1996). When the brain changes its mind: Interocular grouping during binocular rivalry. Proceedings of the National Academy of Sciences, USA, 93 (26), 15508–15511.
Leon, P. S., Vanzetta, I., Masson, G. S., & Perrinet, L. U. (2012). Motion clouds: Model-based stimulus synthesis of natural-like random textures for the study of motion perception. Journal of Neurophysiology, 107 (11), 3217–3226, https://doi.org/10.1152/jn.00737.2011.
Li, H.-H., Carrasco, M., & Heeger, D. J. (2015). Deconstructing interocular suppression: Attention and divisive normalization. PLoS Computational Biology, 11 (10): e1004510, https://doi.org/10.1371/journal.pcbi.1004510.
Ling, S., & Blake, R. (2012). Normalization regulates competition for visual awareness. Neuron, 75 (3), 531–540, https://doi.org/10.1016/j.neuron.2012.05.032.
Livingstone, M. S., & Hubel, D. H. (1987). Psychophysical evidence for separate channels for the perception of form, color, movement, and depth. The Journal of Neuroscience, 7 (11), 3416–3468.
Livingstone, M. S., & Hubel, D. H. (1988, May 6). Segregation of form, color, movement, and depth: Anatomy, physiology, and perception. Science, 240 (4853), 740–749.
Logothetis, N. K., Leopold, D. A., & Sheinberg, D. L. (1996, April 18). What is rivalling during binocular rivalry? Nature, 380 (6575), 621–624, https://doi.org/10.1038/380621a0.
Lorenceau, J., & Alais, D. (2001). Form constraints in motion binding. Nature Neuroscience, 4 (7), 745–751, https://doi.org/10.1038/89543.
Lorenceau, J., & Shiffrar, M. (1992). The influence of terminators on motion integration across space. Vision Research, 32 (2), 263–273, https://doi.org/10.1016/0042-6989(92)90137-8.
Marshak, W., & Sekuler, R. (1979, September 28). Mutual repulsion between moving visual targets. Science, 205 (4413), 1399–1401.
Medathati, N. V. K., Rankin, J., Meso, A. I., Kornprobst, P., & Masson, G. S. (2017). Recurrent network dynamics reconciles visual motion segmentation and integration. Scientific Reports, 7 (1): 11270, https://doi.org/10.1038/s41598-017-11373-z.
Mitchell, J. F., Stoner, G. R., & Reynolds, J. H. (2004, May 27). Object-based attention determines dominance in binocular rivalry. Nature, 429 (6990), 410–413.
Moors, P., Boelens, D., van Overwalle, J., & Wagemans, J. (2016). Scene integration without awareness: No conclusive evidence for processing scene congruency during continuous flash suppression. Psychological Science, 27 (7), 945–956, https://doi.org/10.1177/0956797616642525.
Moors, P., Hesselmann, G., Wagemans, J., & van Ee, R. (2017). Continuous flash suppression: Stimulus fractionation rather than integration. Trends in Cognitive Sciences, 21 (10), 719–721, https://doi.org/10.1016/j.tics.2017.06.005.
Moors, P., Wagemans, J., & de-Wit, L. (2014). Moving stimuli are less effectively masked using traditional continuous flash suppression (CFS) compared to a moving Mondrian mask (MMM): A test case for feature-selective suppression and retinotopic adaptation. PLoS One, 9 (5): e98298, https://doi.org/10.1371/journal.pone.0098298.
Morey, R. D., Rouder, J. N., & Jamil, T. (2014). BayesFactor: Computation of Bayes factors for common designs. R Package Version 0.9, 8.
Nassi, J. J., & Callaway, E. M. (2009). Parallel processing strategies of the primate visual system. Nature Reviews Neuroscience, 10 (5), 360–372, https://doi.org/10.1038/nrn2619.
Nguyen, V. A., Freeman, A. W., & Wenderoth, P. (2001). The depth and selectivity of suppression in binocular rivalry. Perception & Psychophysics, 63 (2), 348–360, https://doi.org/10.3758/BF03194475.
Ooi, T. L., & Loop, M. S. (1994). Visual suppression and its effect upon color and luminance sensitivity. Vision Research, 34 (22), 2997–3003, https://doi.org/10.1016/0042-6989(94)90272-0.
O'Shea, R. P., & Crassini, B. (1981). The sensitivity of binocular rivalry suppression to changes in orientation assessed by reaction-time and forced-choice techniques. Perception, 10 (3), 283–293, https://doi.org/10.1068/p100283.
O'Shea, R. P., & Williams, D. R. (1996). Binocular rivalry with isoluminant stimuli visible only via short-wavelength-sensitive cones. Vision Research, 36 (11), 1561–1571, https://doi.org/10.1016/0042-6989(95)00227-8.
Peirce, J. W. (2007). PsychoPy—Psychophysics software in Python. Journal of Neuroscience Methods, 162 (1–2), 8–13, https://doi.org/10.1016/j.jneumeth.2006.11.017.
Peirce, J. W. (2009). Generating stimuli for neuroscience using PsychoPy. Frontiers in Neuroinformatics, 2: 10, https://doi.org/10.3389/neuro.11.010.2008.
Petrov, Y., Carandini, M., & McKee, S. (2005). Two distinct mechanisms of suppression in human vision. Journal of Neuroscience, 25 (38), 8704–8707, https://doi.org/10.1523/JNEUROSCI.2871-05.2005.
Quaia, C., Optican, L. M., & Cumming, B. G. (2017). Suppression and contrast normalization in motion processing. Journal of Neuroscience, 37 (45), 11051–11066, https://doi.org/10.1523/JNEUROSCI.1572-17.2017.
Quaia, C., Optican, L. M., & Cumming, B. G. (2018). Binocular summation for reflexive eye movements. Journal of Vision, 18 (4): 7, 1–17, https://doi.org/10.1167/18.4.7. [PubMed] [Article]
R Core Team. (2014). R: A language and environment for statistical computing. Retrieved from http://www.R-project.org/.
Rouder, J. N., & Morey, R. D. (2012). Default Bayes factors for model selection in regression. Multivariate Behavioral Research, 47 (6), 877–903, https://doi.org/10.1080/00273171.2012.734737.
Said, C. P., & Heeger, D. J. (2013). A model of binocular rivalry and cross-orientation suppression. PLoS Computational Biology, 9 (3): e1002991, https://doi.org/10.1371/journal.pcbi.1002991.
Smith, E. L., Levi, D. M., Harwerth, R. S., & White, J. M. (1982, November 19). Color vision is altered during the suppression phase of binocular rivalry. Science, 218 (4574), 802–804, https://doi.org/10.1126/science.7134975.
Stanley, J., Forte, J., Cavanagh, P., & Carter, O. (2011). Onset rivalry: The initial dominance phase is independent of ongoing perceptual alternations. Frontiers in Human Neuroscience, 5: 140, https://doi.org/10.3389/fnhum.2011.00140.
Stein, T., Hebart, M. N., & Sterzer, P. (2011). Breaking continuous flash suppression: A new measure of unconscious processing during interocular suppression? Frontiers in Human Neuroscience, 5: 167, https://doi.org/10.3389/fnhum.2011.00167.
Stone, J., Dreher, B., & Leventhal, A. (1979). Hierarchical and parallel mechanisms in the organization of visual cortex. Brain Research Reviews, 1 (3), 345–394, https://doi.org/10.1016/0165-0173(79)90010-9.
Stuit, S. M., Cass, J., Paffen, C. L., & Alais, D. (2009). Orientation-tuned suppression in binocular rivalry reveals general and specific components of rivalry suppression. Journal of Vision, 9 (11): 17, 1–15, https://doi.org/10.1167/9.11.17. [PubMed] [Article]
Tapia, E., & Breitmeyer, B. G. (2011). Visual consciousness revisited: Magnocellular and parvocellular contributions to conscious and nonconscious vision. Psychological Science, 22 (7), 934–942, https://doi.org/10.1177/0956797611413471.
Tsuchiya, N., & Koch, C. (2005). Continuous flash suppression reduces negative afterimages. Nature Neuroscience, 8 (8), 1096–1101.
van de Grind, W. A., van Hof, P., van der Smagt, M. J., & Verstraten, F. A. J. (2001). Slow and fast visual motion channels have independent binocular–rivalry stages. Proceedings of the Royal Society of London B: Biological Sciences, 268 (1465), 437–443, https://doi.org/10.1098/rspb.2000.1380.
Wade, N. J., Weert, C. M. M. D., & Swanston, M. T. (1984). Binocular rivalry with moving patterns. Perception & Psychophysics, 35 (2), 111–122, https://doi.org/10.3758/BF03203891.
Walker, P., & Powell, D. J. (1979). The sensitivity of binocular rivalry to changes in the nondominant stimulus. Vision Research, 19 (3), 247–249, https://doi.org/10.1016/0042-6989(79)90169-X.
Watamaniuk, S. N. J., & Duchon, A. (1992). The human visual system averages speed information. Vision Research, 32 (5), 931–941, https://doi.org/10.1016/0042-6989(92)90036-I.
Wickham, H. (2009). ggplot2: Elegant Graphics for Data Analysis. Retrieved from http://link.springer.com/10.1007/978-0-387-98141-3.
Wiesenfelder, H., & Blake, R. (1990). The neural site of binocular rivalry relative to the analysis of motion in the human visual system. The Journal of Neuroscience, 10 (12), 3880–3888.
Wilson, H. R. (2017). Binocular contrast, stereopsis, and rivalry: Toward a dynamical synthesis. Vision Research, 140, 89–95, https://doi.org/10.1016/j.visres.2017.07.016.
Wolfe, J. M. (1984). Reversing ocular dominance and suppression in a single flash. Vision Research, 24 (5), 471–478.
Yang, E., & Blake, R. (2012). Deconstructing continuous flash suppression. Journal of Vision, 12 (3): 8, 1–14, https://doi.org/10.1167/12.3.8. [PubMed] [Article]
Yang, E., Brascamp, J., Kang, M.-S., & Blake, R. (2014). On the use of continuous flash suppression for the study of visual processing outside of awareness. Consciousness Research, 5: 724, https://doi.org/10.3389/fpsyg.2014.00724.
Zhou, J., Benson, N. C., Kay, K. N., & Winawer, J. (2018). Compressive temporal summation in human visual cortex. Journal of Neuroscience, 38 (3), 691–709, https://doi.org/10.1523/JNEUROSCI.1724-17.2017.
Figure 1
 
An example trial and trial structure. Motion clouds moving at seven different translational velocities (from 0°/s to 16°/s) in opposite directions were presented to two eyes for 1 s. The arrows inside the stimuli schematically show velocity and direction of motion. In this example trial, the stimulus presented to the right eye had a higher velocity. The example percept was that of the slower stimulus in the center and the faster stimulus along the edges (illustrated schematically in the upper-right corner). The participant was to indicate the dominance of the whole field or the central region only if the center differed from the surround. Because the percept (upper right) was center-surround, the participant indicates leftward dominance (i.e., dominance in the center). Feedback was provided with a leftward notch reflecting the response. To indicate that leftward motion was only dominant in the center, the participant adjusted the size of the red circle to approximately reflect the area of the center. In this case, the center size was indicated to be ∼75% of the stimulus diameter. Note that the faster stimulus (right) introduces motion streaks due to high velocity; these are eliminated in further experiments by using temporal frequency modulation instead of directional motion (see Supplementary Files S1S5 for stimulus examples).
Figure 1
 
An example trial and trial structure. Motion clouds moving at seven different translational velocities (from 0°/s to 16°/s) in opposite directions were presented to two eyes for 1 s. The arrows inside the stimuli schematically show velocity and direction of motion. In this example trial, the stimulus presented to the right eye had a higher velocity. The example percept was that of the slower stimulus in the center and the faster stimulus along the edges (illustrated schematically in the upper-right corner). The participant was to indicate the dominance of the whole field or the central region only if the center differed from the surround. Because the percept (upper right) was center-surround, the participant indicates leftward dominance (i.e., dominance in the center). Feedback was provided with a leftward notch reflecting the response. To indicate that leftward motion was only dominant in the center, the participant adjusted the size of the red circle to approximately reflect the area of the center. In this case, the center size was indicated to be ∼75% of the stimulus diameter. Note that the faster stimulus (right) introduces motion streaks due to high velocity; these are eliminated in further experiments by using temporal frequency modulation instead of directional motion (see Supplementary Files S1S5 for stimulus examples).
Figure 2
 
Experiment 1: Trials reported as stationary-dominant or center-surround. (A) The proportion of stationary-dominant (both whole-field and center-surround) trials. Top: Stationary-dominant trial proportion plotted as a function of the lower (x-axis) and higher (y-axis) of the two competing velocities. Bottom: Same trials plotted as a function of the higher velocity (x-axis). When both velocities were sufficiently low, stationary dominance was most frequent (top graph: lower-left corner; bottom graph: left side of the plot). Interestingly, the slow stimulus also dominated on a majority of trials when competing with a high-velocity stimulus (top graph: upper-left corner; bottom graph: right side of the plot). When the higher velocity was medium (e.g., 4°/s in bottom graph), stationary dominance was less likely to occur. (B) The number of whole-field trials given stationary dominance. These trials occurred more often when both velocities were low (top graph: bottom rows). (C) The number of center-surround trials given stationary dominance. Such trials were closely associated with high-versus-low velocity competition (top graph: upper rows). Dots in the top graphs are group averages. Bottom graphs show standard error (N = 5).
Figure 2
 
Experiment 1: Trials reported as stationary-dominant or center-surround. (A) The proportion of stationary-dominant (both whole-field and center-surround) trials. Top: Stationary-dominant trial proportion plotted as a function of the lower (x-axis) and higher (y-axis) of the two competing velocities. Bottom: Same trials plotted as a function of the higher velocity (x-axis). When both velocities were sufficiently low, stationary dominance was most frequent (top graph: lower-left corner; bottom graph: left side of the plot). Interestingly, the slow stimulus also dominated on a majority of trials when competing with a high-velocity stimulus (top graph: upper-left corner; bottom graph: right side of the plot). When the higher velocity was medium (e.g., 4°/s in bottom graph), stationary dominance was less likely to occur. (B) The number of whole-field trials given stationary dominance. These trials occurred more often when both velocities were low (top graph: bottom rows). (C) The number of center-surround trials given stationary dominance. Such trials were closely associated with high-versus-low velocity competition (top graph: upper rows). Dots in the top graphs are group averages. Bottom graphs show standard error (N = 5).
Figure 3
 
Experiment 2: The effect of stimulus size on motion dominance. (A) Proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent dominance responses as a function of stimulus size. The larger stimuli were more likely to produce stationary dominance. (B) The incidence of center-surround segmentation (orange curve) and the size of the stationary center (blue curve) as a function of stimulus size. The majority of the stationary-dominant trials had center-surround pattern. The size of the center was a constant proportion of the stimulus size. Dots are group averages; error bars reflect standard errors. See Supplementary Figure S4 for individual data (N = 5).
Figure 3
 
Experiment 2: The effect of stimulus size on motion dominance. (A) Proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent dominance responses as a function of stimulus size. The larger stimuli were more likely to produce stationary dominance. (B) The incidence of center-surround segmentation (orange curve) and the size of the stationary center (blue curve) as a function of stimulus size. The majority of the stationary-dominant trials had center-surround pattern. The size of the center was a constant proportion of the stimulus size. Dots are group averages; error bars reflect standard errors. See Supplementary Figure S4 for individual data (N = 5).
Figure 4
 
Experiment 3: The effect of stationary and dynamic stimulus delay on dominance. The leftmost graph is the average. (A) The proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent responses as a function of stationary stimulus delay. Varying the stationary stimulus asynchrony did not change the incidence of stationary dominance for the following second. (B) Proportion of different response types as a function of dynamic stimulus delay following stationary stimulus. Introducing dynamic stimulus >600 ms after stationary stimulus largely abolished stationary dominance for most participants in favor of piecemeal rivalry. Dots are group averages; error bars reflect standard errors.
Figure 4
 
Experiment 3: The effect of stationary and dynamic stimulus delay on dominance. The leftmost graph is the average. (A) The proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent responses as a function of stationary stimulus delay. Varying the stationary stimulus asynchrony did not change the incidence of stationary dominance for the following second. (B) Proportion of different response types as a function of dynamic stimulus delay following stationary stimulus. Introducing dynamic stimulus >600 ms after stationary stimulus largely abolished stationary dominance for most participants in favor of piecemeal rivalry. Dots are group averages; error bars reflect standard errors.
Figure 5
 
Experiment 4: The effect of trial duration on motion dominance and center size. (A) Proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent responses as a function of trial duration. Trials up to 1 s in duration were more likely to result in stationary dominance at the end of the trial. (B) The proportion of center/surround trials among stationary-dominant trials (in orange) and the size of the stationary center (in blue) as a function of trial duration. The stationary stimulus dominated in the center but not the periphery on a majority of trials. As the trials increased in duration, the size of the stationary center decreased slightly. The size of the center is expressed as a proportion relative to the stimulus diameter. Dots are group averages; error bars reflect standard errors. See Supplemental Figure S5 for individual data (N = 8).
Figure 5
 
Experiment 4: The effect of trial duration on motion dominance and center size. (A) Proportion of stationary-dominant, dynamic-dominant, piecemeal, and transparent responses as a function of trial duration. Trials up to 1 s in duration were more likely to result in stationary dominance at the end of the trial. (B) The proportion of center/surround trials among stationary-dominant trials (in orange) and the size of the stationary center (in blue) as a function of trial duration. The stationary stimulus dominated in the center but not the periphery on a majority of trials. As the trials increased in duration, the size of the stationary center decreased slightly. The size of the center is expressed as a proportion relative to the stimulus diameter. Dots are group averages; error bars reflect standard errors. See Supplemental Figure S5 for individual data (N = 8).
Figure 6
 
An example trial in Experiment 5. Each trial started with a simultaneous onset of a mask to the dominant eye and a target to the nondominant eye. Mask speed bandwidth was randomly assigned to 1°/s, 2°/s, 4°/s, or 8°/s, and the target was either stationary or fast (0°/s or 16°/s, respectively). In this example trial, the target was “near” at 0.9° off-center; a “far” target would be presented closer to the edge at 2.7° off-center.
Figure 6
 
An example trial in Experiment 5. Each trial started with a simultaneous onset of a mask to the dominant eye and a target to the nondominant eye. Mask speed bandwidth was randomly assigned to 1°/s, 2°/s, 4°/s, or 8°/s, and the target was either stationary or fast (0°/s or 16°/s, respectively). In this example trial, the target was “near” at 0.9° off-center; a “far” target would be presented closer to the edge at 2.7° off-center.
Figure 7
 
Contrast detection threshold as a function of mask speed, target speed, and target eccentricity. Stationary targets were more effectively suppressed by slower masks and vice versa for fast targets.
Figure 7
 
Contrast detection threshold as a function of mask speed, target speed, and target eccentricity. Stationary targets were more effectively suppressed by slower masks and vice versa for fast targets.
Table 1
 
The percentage of trials of different dominance judgments in terms of the velocity (column) and spatial distribution (rows). Notes: Trials with both velocities being equal (2.5% of trials) were not analyzed for responses. The standard deviation is in parentheses. The italicized cells indicate dominance categories omitted from further analysis due to low incidence (<5% of all trials).
Table 1
 
The percentage of trials of different dominance judgments in terms of the velocity (column) and spatial distribution (rows). Notes: Trials with both velocities being equal (2.5% of trials) were not analyzed for responses. The standard deviation is in parentheses. The italicized cells indicate dominance categories omitted from further analysis due to low incidence (<5% of all trials).
Table 2
 
Regression and Bayes factor results of the mask speed, target speed, and target eccentricity. Notes: P and BF values in bold are significant or approaching significance. Higher BF values indicate support of the model that includes the corresponding variable (nominator model) compared to the model that excludes it (denominator model).
Table 2
 
Regression and Bayes factor results of the mask speed, target speed, and target eccentricity. Notes: P and BF values in bold are significant or approaching significance. Higher BF values indicate support of the model that includes the corresponding variable (nominator model) compared to the model that excludes it (denominator model).
Supplement 1
Supplement 2
Supplement 3
Supplement 4
Supplement 5
Supplement 6
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×