Color-binding errors induced by modulating effects of the preceding stimulus on onset rivalry

Abstract

Onset rivalry can be modulated by a preceding stimulus with features similar to rivalrous test stimuli. In this study, we used this modulating effect to investigate the integration of color and orientation during onset rivalry using equiluminant chromatic gratings. Specifically, we explored whether this modulating effect leads to a decoupling of color and orientation in chromatic gratings, resulting in a percept distinct from either of the rivalrous gratings. The results demonstrated that color-binding errors can be observed in a form where rivalrous green–gray clockwise and red–gray counterclockwise gratings yield the percept of a bichromatic, red–green grating with either clockwise or counterclockwise orientation. These errors were observed under a brief test duration (30 ms), with both monocular and binocular presentations of the preceding stimulus. The specific color and orientation combination of the preceding stimulus was not critical for inducing color-binding errors, provided it was composed of the test color and orientation. We also found a notable covariant relationship between the perception of color-binding errors and exclusive dominance, where the perceived orientation in color-binding errors generally matched that in exclusive dominance. This finding suggests that the mechanisms underlying color-binding errors may be related to, or partially overlap with, those determining exclusive dominance. These errors can be explained by the decoupling of color and orientation in the representation of the suppressed grating, with the color binding to the dominant grating, resulting in an erroneously perceived bichromatic grating.

Introduction

Binocular rivalry is a perceptual conflict that occurs when the images projected to each eye are sufficiently different. During binocular rivalry, observers perceive only one image at a time (exclusive dominance), with the dominant image switching from one to the other unpredictably (Levelt, 1965). Alternatively, observers may perceive a mosaic of two images (piecemeal rivalry). Because the percept during rivalry fluctuates even though the images presented to both eyes remain unchanged, binocular rivalry has been the subject of studies investigating perceptual awareness and the mechanism of binocular integration. The prevailing theory regarding the visual mechanism mediating binocular rivalry suggests a hybrid model. This model posits that competitive interactions at multiple levels of the processing hierarchy contribute to the determination of dominance and suppression during rivalry (Blake & Logothetis, 2002; Wilson, 2003; Tong, Meng, & Blake, 2006). These competitive interactions involve two main aspects: eye-based and feature-based competitions. Eye-based competition corresponds with the competition between monocular processes that receive incompatible images from the left and right eyes, whereas feature-based competition refers to the competition between incompatible representations of stimulus features comprising rivalrous images. 

Exploring the feature-based competition that underlies binocular rivalry enhances our understanding of how stimulus features are integrated binocularly and how a unitary percept is formed through integrating different features. One approach to this investigation is to elucidate whether conflicts involving multiple stimulus features are resolved collectively, with all features being combined, or individually, with each feature being resolved separately. Previous studies have identified three types of perception that can occur in multifeature integration, depending on stimulus conditions and experimental procedures: exclusive dominance, misbinding, and color-binding errors. Exclusive dominance refers to the perception wherein only one of the rivalrous stimuli is perceived. As previously described, it is commonly observed during binocular rivalry, whereas the other two types are reported less frequently. Misbinding occurs when different visual features from the rivalrous stimuli compete separately, resulting in a final percept that differs from either of the original stimuli (Andrews & Blakemore, 2002; Holmes, Hancock, & Andrews, 2006). For instance, Holmes et al. (2006) presented rivalrous stimuli differing in both color and orientation, showing that pattern rivalry could be resolved independently of color rivalry. They revealed that observers occasionally reported perceiving a chromatic stimulus with the color from one rivalrous stimulus and the orientation from the other when the stimuli were presented immediately after a binocular preceding stimulus. Misbinding occurred when the preceding stimulus shared the same color as one rivalrous stimulus and the same pattern as the other, indicating that the binocular preceding stimulus modulated color and pattern rivalry separately. 

Color-binding errors refer to perceptions in which, for instance, rivalrous red–gray horizontal and green–gray vertical gratings yield the percept of a bichromatic, red–green horizontal grating (Hong & Shevell, 2006; Kang & Shevell, 2008; Maloney, Lam, & Clifford, 2013). In this type of perception, only one of the rivalrous orientations, horizontal or vertical, becomes dominant. Still, the color of the suppressed grating is incorrectly associated with the gray region of the dominant grating. Hong and Shevell (2006) reported that color-binding errors occur when rivalrous chromatic gratings possess low luminance contrast (<20% in Michelson contrast) and have been observed for a prolonged period (>30 seconds). 

In investigating how different stimulus features are combined and how the three types of perception occur in binocular rivalry, the initial dominance modulation (IDM) paradigm can serve as a valuable tool. The IDM paradigm is an experimental procedure where a preceding stimulus is presented before the rivalrous stimuli, with a variable interstimulus interval (ISI) between the preceding and rivalrous stimuli (Brascamp, Knapen, Kanai, van Ee, & van den Berg, 2007; Abe, Kimura, & Goryo, 2011). Pre-exposure to a stimulus can modulate the initial selection of dominance for rivalrous stimuli (onset rivalry), both suppressively and facilitatively. Previous studies demonstrated that a brief preceding stimulus with low contrast had facilitative effects, while a longer preceding stimulus with higher contrast tended to have suppressive effects (Brascamp et al., 2007; Abe et al., 2011). Moreover, when tested with a longer preceding stimulus (e.g., 1 second), eye-based suppression was observed with shorter test durations (e.g., 10 ms) and shorter ISIs. In contrast, feature-based suppression became more pronounced with longer test durations (e.g., 200 ms) and longer ISIs (Abe et al., 2011). 

It should be noted that the term “initial dominance modulation” paradigm has been coined to distinguish it from “flash suppression” (Abe et al., 2011). The latter term more accurately describes a procedure wherein a half-image of rivalrous stimuli is initially presented to one eye in isolation, and subsequently, the other half-image is shown to the opposite eye without an ISI (Wolfe, 1984; Ooi & Loop, 1994). In this procedure, the sudden onset (i.e., flash) of the second half-image generates a transient signal in only one eye, which can immediately suppress the perception of the first half-image. Conversely, in the IDM paradigm, which includes an ISI, the focus is more on how a preceding stimulus modulates perceptual dominance when the two rivalrous stimuli are simultaneously presented. 

Using the IDM paradigm, Abe et al. (2011) investigated whether feature-based suppression operates independently for color and orientation in rivalrous chromatic gratings with high luminance contrast. They found that the selectivity to feature combination was manifested differently in the modulating effect of the preceding stimulus and in the competitive interaction for perceptual dominance. The color and orientation of the preceding stimulus almost independently modulated the dominance of the rivalrous stimuli. For example, previewing binocularly presented color disks or achromatic gratings similarly suppressed one of the rivalrous chromatic gratings possessing the same feature. However, competitive interactions for perceptual dominance were resolved in a manner selective for feature combination. Even when a preceding stimulus was composed of a different color–orientation combination from the rivalrous test stimuli (e.g., a green grating tilted 45° from vertical was used as a preceding stimulus when the rivalrous stimuli were a green grating tilted 135° and a red grating tilted 45°), the resulting percept corresponded to one of the rivalrous chromatic gratings (i.e., exclusive dominance). In their study, misbinding between color and orientation and color-binding errors were never reported. 

The objective of this study was to extend the previous investigation using the IDM paradigm, with the aim of examining multifeature integration between color and orientation using equiluminant chromatic gratings. Specifically, we explored whether and under what stimulus conditions pre-exposure to a preceding stimulus might lead to misbinding or color-binding errors. Hong and Shevell (2006) proposed that luminance edges are pivotal in maintaining the correct binding of color and form at the right region of the stimuli during sustained rivalry. If this also applies to onset rivalry, the color-binding errors might also be observed with equiluminant gratings in the IDM paradigm. We systematically varied the features of the preceding stimulus in relation to those of the rivalrous chromatic gratings, as well as the timing of the stimulus presentation, including the duration of the rivalrous stimuli and the ISI. Extending Abe et al. (2011), we examined the effects of the monocular as well as the binocular preceding stimulus. We tested the color- and orientation-congruent conditions using the preceding stimulus composed of an equiluminant chromatic grating instead of the color-disk and achromatic-grating conditions used in Abe et al. (2011) (see Methods). 

We anticipated that misbinding or color-binding errors could be found in the stimulus conditions where feature-based suppression has been shown to be predominant over eye-based suppression, e.g., when the test duration was long (e.g., 200 ms) (Abe et al., 2011). However, in contrast with our expectations, color-binding errors were found when the test duration was short and under conditions where competitive interactions could be resolved in an eye-based manner. 

Methods

Apparatus and stimulus

The apparatus was the same haploscope used in our previous study (Abe et al., 2011), which consisted of three 17-inch color CRT monitors and two beam splitters. One monitor (PressView17SR, Radius Inc., Sunnyvale, CA) was positioned in front of the observer, on which a fixation pattern was presented. The observer viewed the fixation pattern with natural accommodation and convergence. The other two monitors (Flex Scan T566, EIZO Corporation, Ishikawa, Japan) were positioned on the left and right sides of the observer, on which the stimuli for the left and right eye were presented, respectively. The stimulus images on the left and right monitors were superimposed on the image on the front monitor using the beam splitters. The front monitor was controlled with a Radius board, with a pixel resolution of 1,024 × 768 and a frame rate of 75 Hz. The left and right monitors were controlled using two VSG 2/5 graphic cards with a 15-bit color resolution (Cambridge Research Systems, Ltd., Kent, UK) and Matlab (The MathWorks Inc., Natick, MA). The two monitors had a pixel resolution of 1,024 × 768 and a frame rate of 100 Hz. Their refresh timing was synchronized using the built-in functionality of VSG cards. The Psychophysics Toolbox extensions for Matlab were used in the phosphor calibrations and colorimetric calculations (Brainard, 1997; Pelli, 1997; Kleiner, Brainard, & Pelli, 2007). The optical distance was 57 cm from the eyes. The observer's head was stabilized by a chin and forehead rest. 

Both preceding and rivalrous test stimuli were 2° circular fields located at the center of a dark gray background (x = 0.313, y = 0.329). The background subtended 10° × 10°, and its luminance was 0.1 cd/m². Test stimuli were a pair of equiluminant chromatic gratings differing in color and orientation. Each test stimulus was a 2 cpd square-wave grating composed of color and gray stripes, with a mean luminance of 4 cd/m². The color stripes were red (x = 0.378, y = 0.298) or green (x = 0.232, y = 0.368), and the gray stripes were the same gray as the background (x = 0.313, y = 0.329). The color was chosen along the ± (L–M) direction in DKL color space (Derrington, Krauskopf, & Lennie, 1984). The color contrast was set to the highest contrast available for one color, and the same value was used for the other color. Throughout all experiments, the test stimuli were always composed of a green–gray clockwise-tilted (green CW) grating presented to the left eye and a red–gray counterclockwise-tilted (red CCW) grating presented to the right eye. 

Stimulus condition

To investigate the effects of the preceding color, orientation, and their combination on the resolution of onset rivalry of the test stimuli, we manipulated the preceding stimulus in the following four conditions: same combination, different combination, color-congruent, and orientation-congruent conditions. The preceding stimulus was always composed of a chromatic grating. In the same combination condition, the preceding stimulus was identical to one of the test stimuli (i.e., a green CW or a red CCW grating). In the different combination condition, the preceding stimulus was composed of the color from one of the test stimuli and the orientation of the other (i.e., a red CW or a green CCW grating). In the color-congruent condition, the preceding stimulus was a chromatic grating that shared only color with one of the test stimuli (i.e., a green–gray or red–gray grating with vertical orientation). In the orientation-congruent condition, the preceding stimulus was a chromatic grating that shared only orientation with one of the test stimuli (i.e., a blue–gray CW or CCW grating). 

We also manipulated the eye of presentation of the preceding stimulus in the binocular and monocular presentation conditions. In the binocular condition, the preceding stimulus was presented to both eyes to investigate the effects of stimulus features of the preceding stimulus. In the monocular condition, the preceding stimulus was presented to either eye to investigate the effects of the presented eye. Thus, there were 24 varieties of the preceding stimulus (8 color–orientation combinations were crossed with 3 varieties of the presented eye, i.e., binocular, left, and right eye). 

Observers

Four observers participated in the same and different combination conditions. One of the observers (O₄) was not available for the color- and orientation-congruent conditions. One observer was the first author, and the others were naïve regarding the purpose of the experiment. All observers had normal or corrected-to-normal visual acuity, good stereopsis, and normal color vision as assessed with the Ishihara pseudo-isochromatic plates. Before the experiment, all observers provided written informed consent after a thorough explanation of the procedures. The experiments were conducted in accordance with the Declaration of Helsinki. 

Procedure

In each trial, a preceding stimulus was presented for 1000 ms to one eye or both eyes, depending upon the preceding stimulus conditions. After an ISI of 20, 50, or 200 ms, test stimuli were dichoptically presented to two eyes. The duration of the test stimuli was 30 or 200 ms. The temporal condition was determined based on our previous findings (Abe et al., 2011). The shorter test duration was set to 30 ms to ensure reliable orientation judgments with equiluminant gratings. In the control condition, only the test stimuli were presented with the same timing as in the other stimulus conditions. 

The observer's task was to report the appearance of the test stimuli. Because the test stimuli were equiluminant gratings and the test duration was rather short, the task was difficult. However, all observers had good experience in similar types of binocular rivalry experiments. Before the experiment, the observers were informed that the test stimuli were composed of a green CW and a red CCW grating presented dichoptically, and they observed each monocular grating separately. They also observed the dichoptically presented test stimuli of long duration without a preceding stimulus to experience piecemeal percepts (i.e., a mosaic-like appearance consisting of patches of each test stimulus from one eye or the other). Then, they were instructed to report a piecemeal percept if the mosaic-like appearance was found in even a small part of the stimulus field. They were also instructed to report the color and orientation of a grating or other spatial pattern such as a plaid in their percept. 

After the measurements, the observer's report was classified into representative categories. It was found that three categories, that is, exclusive dominance, color-binding error, and piecemeal percept, were necessary to classify almost all the results. Exclusive dominance meant that observers perceived one of the test stimuli alone (i.e., a green CW or red CCW grating). The color-binding error indicated that observers perceived a bichromatic, red–green CW or CCW grating. Two stimulus colors, each presented to the left or right eye, were bound with one orientation (CW or CCW) of the test stimuli, resulting in a red–green alternating grating. (In other words, the gray stripes of the dominant grating took on the color of the suppressed grating presented to the different eye.) Other misbound percepts were rarely reported; thus, only the red–green CW or CCW grating was classified in the color-binding error category. Appearances other than the three categories were seldom reported and thus excluded from further analyses. We categorized the observer's report as piecemeal only when the patches were composed of test gratings, that is, a green CW and a red CCW grating. Otherwise, we categorized a patchy appearance as appearances other than the three categories. The proportion of excluded trials was 0.002 for observer O₁, 0.0 for O₂, 0.069 for O₃, and 0.093 for O₄. In almost all of these irregular cases, observers reported patchy appearances but could not identify the appearance of some patches confidently. 

Experiments were conducted in a dark room. Before the experiment, the observers were dark adapted for 5 minutes and then light adapted to the background field for 2 minutes. 

Because we manipulated the preceding stimulus conditions (e.g., color-congruent condition), eye of presentation (e.g., binocular), the combination of the preceding color and orientation (e.g., green vertical), and ISI (e.g., 20 ms), the description of how we tested these different conditions is rather complicated. To clarify, we prepared condition tables (Supplementary Tables S1 and S2) listing the conditions and the number of trials for each. In the experiment, we tested the color- and orientation-congruent conditions separately from the same and different combination conditions, as shown in Supplementary Tables S1 and S2, respectively. For each list of conditions described in the Supplementary Tables, we assigned four trial blocks for the short test duration (30 ms) and another four trial blocks for the long test duration (200 ms). The short and long test durations were tested in different daily sessions, with each session dedicated to either the short or long duration. Each daily session included two trial blocks, requiring a total of four daily sessions per test duration to complete all measurements. 

Each trial block included 12 variations of the preceding stimulus conditions (two color–orientation combinations crossed with three varieties of the presented eye for two different preceding stimulus conditions) and control conditions (Supplementary Tables S1 and S2). These were tested four times for each of three different ISI conditions and randomly interleaved, resulting in a total of 156 trials per block. Each combination of conditions was tested 16 times for each observer. For the same preceding stimulus conditions, different daily sessions were repeated within a 1- or 2-week period. The control condition was tested in both the trial blocks for the color- and orientation-congruent conditions and those for the same and different combination conditions (48 trials each). For the three observers who participated in all preceding stimulus conditions, the data from all trial blocks were pooled for the control condition. 

Data analysis

Because this study was primarily exploratory rather than hypothesis testing, we repeated measurements for a small group of well-experienced observers to obtain reliable data, and the effect of a preceding stimulus was analyzed at the individual observer level. Our approach examined whether the presentation of the preceding stimulus increased or decreased the proportion of specific appearance categories, such as exclusive dominance and color-binding errors, compared with the control condition, where the rivalrous test stimuli were presented alone. This comparison was made using the bootstrap method to calculate the 95% confidence interval for the proportion of appearance categories in the control condition and then examining whether the modulating effect of the preceding stimulus caused the changes in the proportion to fall outside this interval. The confidence interval was calculated using the bias-corrected and accelerated method based on 9,999 bootstrap samples (each sample consisted of 48 trials resampled with replacement) with the scipy.stats.bootstrap python module (Efron & Tibshirani, 1994; Virtanen et al., 2020). 

Throughout all experiments, color-binding errors were never reported by any of the observers in the control condition. Thus, the 95% confidence interval for the results of color-binding errors was not shown in the figures. 

Results

In each condition of the preceding stimulus, the proportions of each appearance category (i.e., exclusive dominance, color-binding error, and piecemeal rivalry) were calculated for each observer. The results for different ISIs were pooled within each observer because consistent tendencies were not found across observers (48 trials in total for each variation of the preceding stimulus). In the following sections, the results for the test duration of 30 ms will be reported because color-binding errors were rarely reported when the test duration was 200 ms. The results for the 200 ms duration were consistent with previous findings (Abe et al., 2011) and are described in the Supplementary Results. 

Results in the color- and orientation-congruent conditions

Figures 1b and 2b show the proportion of exclusive dominance in the color- and orientation-congruent conditions, respectively. The specific stimulus sequence in each condition is illustrated in Figures 1a and 2a. Color-binding errors were rarely reported in these conditions, and the total proportions of the color binding errors for individual observers are shown in the Supplementary Results (Table S1). 

Figure 1.

 

Stimulus sequence and results in the color-congruent condition for a test duration of 30 ms. (a) The specific stimulus sequence of the preceding and test stimuli is illustrated for both the binocular and monocular presentation of the preceding stimulus. (b) The proportions of exclusive dominance. The results for feature-based modulation are shown in the left panel, while those for eye-based modulation are in the right panel. Different symbols represent the results for individual observers. White and gray symbols in the feature-based modulation panel represent different and same color dominance, respectively, whereas white and gray symbols in the eye-based modulation panel represent contralateral and ipsilateral dominance, respectively (see text for the explanation of how the proportion of each dominance was calculated). Error bars in Figure 1b show the 95% bootstrap confidence interval of exclusive dominance in the control condition. In the color- and orientation-congruent conditions (Figures 1b and 2b), the proportion of exclusive dominance in the control condition was zero for O₁ and O₂ (circles and triangles, respectively) but 0.30 for O₃ (squares). White and gray bars show the mean proportion of exclusive dominance averaged across observers.

Figure 2.

 

Stimulus sequence (a) and results (b) in the orientation-congruent condition for a test duration of 30 ms. In Figure 2b, white and gray symbols in the feature-based modulation panel (left) represent different and same orientation dominance, respectively (see text for explanation). Other aspects are the same as those in Figure 1.

Presentation of the results

In these and subsequent figures, different white and gray symbols illustrate the results for different observers, and white and gray bars show the results averaged across observers. Each unique symbol represents a specific observer. The results were organized in terms of feature-based and eye-based modulations (left and right panels in Figures 1b and 2b, respectively). To organize the results, the data were averaged across pairs of symmetrical situations. For example, when the preceding stimulus was presented binocularly in the color-congruent conditions (Figure 1a), the proportion of the red CCW percept produced by the green vertical preceding grating was averaged with that of the green CW percept produced by the red preceding grating, resulting in the proportion of exclusive dominance with a different color (hereafter referred to as different color dominance), shown with white symbols (Figure 1b, left). Similarly, the proportion of exclusive dominance with the same color (same color dominance) was calculated by averaging the proportion of the red CCW percept produced by the red preceding stimulus and that of the green CW percept produced by the green preceding stimulus, shown with gray symbols (Figure 1b, left). In the orientation-congruent condition, white and gray symbols in the feature-based modulation panel (Figure 2b, left) represent different and same orientation dominance, respectively. 

The results for the monocular presentation were shown in both feature-based and eye-based fashions. The feature-based results were derived in the same manner as those for binocular presentation. The eye-based results were summarized as the proportion of exclusive dominance of the grating presented to the eye contralateral to the preceding stimulus (hereafter referred to as contralateral dominance), shown with white symbols, and the proportion of exclusive dominance of the grating presented to the ipsilateral eye (ipsilateral dominance), shown with gray symbols (Figures 1b and 2b, right). The feature-based and eye-based presentations of the results make it possible to examine how the preceding stimulus modulated the dominant percept of the rivalrous test stimuli, decreasing the influences of biases specific to a particular eye, color, and/or orientation. 

For the results in the control condition, the proportions of the exclusively dominant green CW percept and the red CCW percept were averaged to derive the proportion of exclusive dominance, and the 95% bootstrap confidence intervals (see the Data analysis section in the Methods) are shown by error bars in the results for different and same color dominance in Figure 1b and those for different and same orientation dominance in Figure 2b. Because the eye of presentation of rivalrous test stimuli was fixed in this study, the confidence intervals are common for the feature-based (Figures 1b and 2b, left) and eye-based (Figures 1b and 2b, right) modulations. 

Modulating effects in the color- and orientation-congruent conditions

When the red or green vertical preceding stimulus was presented binocularly in the color-congruent condition, the proportion of different color dominance increased (i.e., white symbols are located above the confidence interval shown by error bars), whereas the proportion of same color dominance did not (Figure 1b, left). Thus, the exclusive dominance was determined mainly by color-based suppression. In contrast, when the preceding stimulus was presented monocularly, the results were consistent with combined effects of color- and eye-based suppression. When the results were shown as feature-based modulation (Figure 1b, left), the proportion of different color dominance was high (white symbols). However, when shown as eye-based modulation (Figure 1b, right), the proportion of contralateral dominance was high (white symbols). Thus, both feature- and eye-based suppressions seem to contribute to the resolution of onset rivalry. 

Figure 2b shows the results of exclusive dominance in the orientation-congruent condition. When the preceding stimulus was presented binocularly (Figure 2b, left), exclusive dominance was mostly determined by orientation-based facilitation, although the effects were relatively small (gray symbols). However, when the results for the monocular presentation were presented as feature-based modulation (Figure 2b, left), the proportions of same and different orientation dominance were nearly the same. In contrast, when the results were presented as eye-based modulation (Figure 2b, right), a strong eye-based suppression was observed. The proportion of contralateral dominance (white symbols) increased compared with the proportion of exclusive dominance in the control condition. The results for the monocular presentation can be explained by only assuming eye-based suppression; that is, pre-exposure to a monocular blue CW or CCW grating differentially increased contralateral dominance compared with ipsilateral dominance. Because the contralateral grating could have the same or different orientation relative to the preceding stimulus with equal probability, the proportions of same and different orientation dominance increased similarly. 

The results for the binocular preceding stimulus in the color- and orientation-congruent conditions suggested that different features of the preceding stimulus could modulate onset rivalry separately. Even when only the color or orientation of the preceding stimulus was shared with one of the rivalrous test stimuli, the binocular preceding stimulus could modulate the visibility of the test stimulus in a feature-specific manner. In contrast, in both the color- and orientation-congruent conditions, color-binding errors were only rarely reported, and one observer (O₃) did not report them at all (Table S1 in the Supplementary Results). 

Modulating effects based on the combination of preceding color and orientation

Figures 3 and 4 show the stimulus sequence and results in the same and different combination conditions, respectively. In these conditions, color-binding errors were reported more frequently by all observers, compared with the color- and orientation-congruent conditions. 

Figure 3.

 

Stimulus sequence (a) and results (b and c) in the same combination condition for a test duration of 30 ms. In Figures 3 and 4, the results of color-binding errors (c) are shown in addition to those of exclusive dominance (b). In Figures 3b and 3c, the percepts represented by white and gray symbols in each panel are shown in the legend. In the same and different combination conditions (Figures 3b and 4b), the proportion of exclusive dominance in the control condition was zero for observer O₁ (circles), 0.02 for O₂ (triangles), 0.28 for O₃ (squares), and 0.03 for O₄ (diamonds). Other aspects are the same as those in Figure 1.

Figure 4.

 

Stimulus sequence (a) and results (b and c) in the different combination condition for a test duration of 30 ms. Other aspects are the same as those in Figure 3.

Exclusive dominance caused by the binocular preceding stimulus

If the combination of color and orientation in the binocular preceding stimulus was critical for modulating the visibility of the test stimulus, a strong feature-based modulating effect should have been observed in the same combination condition. For example, the binocular green CW preceding stimulus should have either decreased or increased the proportion of exclusive dominance of the green CW test grating (referred to as same orientation and same color dominance). However, this prediction was not confirmed, and the modulating effects varied among observers (Figure 3b, left). Upon examining the results for each observer, it was found that the binocular preceding stimulus increased both same orientation and same color dominance and different orientation and different color dominance for observer O₁ (represented by circles), increased different orientation and different color dominance for O₂ (triangles), increased same orientation and same color dominance for O₃ (squares), and did not have much of an effect for O₄ (diamonds). 

In contrast, strong differential modulating effects were observed in the different combination condition (Figure 4b, left). The binocular preceding stimulus increased the proportion of same orientation and different color dominance for all observers. These results could be explained by considering the possibility that the preceding color and orientation modulated the visibility of the test stimulus separately but jointly through color-based suppression and orientation-based facilitation, as observed in the color- and orientation-congruent conditions, respectively (Figures 1b and 2b). In the same combination condition, the color-based suppression would counteract the orientation-based facilitation, resulting in a weak or inconsistent modulating effect. Conversely, these two modulating effects would work synergistically in the different combination condition, resulting in a stronger and more consistent modulating effect. 

Exclusive dominance caused by the monocular preceding stimulus

When the preceding stimulus was presented monocularly in the same combination condition, the feature-based modulating effects were also inconsistent across observers (Figure 3b, left). However, the eye-based suppressive effects were strong and more consistent (Figure 3b, right). The proportion of contralateral dominance (white symbols) increased after exposure to a monocular preceding stimulus for all observers. In contrast, in the different combination condition, feature-based same orientation and different color dominance (white symbols) was somewhat more pronounced than different orientation and same color dominance (gray symbols). In contrast, eye-based modulation was generally weak and inconsistent across observers, although contralateral dominance was observed for one observer (circles). 

Here, the effects of the preceding stimulus can be understood by considering a monocular-specific modulating effect (i.e., eye-based suppression), as well as the modulating effects found for the binocular presentation (i.e., color-based suppression and orientation-based facilitation). As described for the binocular presentation elsewhere in this article, in the same combination condition, color-based and orientation-based modulating effects counteracted each other. Consequently, the resulting effects related to specific features were weak or varied among observers, making eye-based suppression relatively more pronounced, with contralateral dominance being reported more frequently. In contrast, in the different combination condition, color-based and orientation-based effects were synergistic, resulting in a more pronounced feature-based effect compared with the eye-based effect. 

Color-binding errors

Color-binding errors were reported by all observers in both the same and different combination conditions (Figures 3c and 4c). When the error was reported, the appearance of the chromatic grating was not patchy, and red–green spatial alternation was evident in the perceived grating. Because the error was rarely reported in the color- and orientation-congruent conditions (Table S1 in the Supplementary Results), it seems to be important for the preceding stimulus to have one color (red or green) and one orientation (CW or CCW) of the test stimuli to induce color-binding errors reliably. However, the specific combination of the preceding color and orientation was not critical. The error was reported consistently in both the same and different combination conditions. Notably, the eye of presentation was also not critical for the error to be reliably induced. The error was observed regardless of whether the preceding stimulus was presented binocularly or monocularly. 

Another notable aspect of the results was that the perceived orientation in color-binding errors generally covaried with that in exclusive dominance (compare the results for exclusive dominance and those for color-binding errors in Figures 3 and 4). When one of the test gratings was reported more frequently as being exclusively dominant (e.g., the results for the binocular presentation in the different combination condition; Figure 4b, left), the grating having the same orientation but with red–green color alternation was reported more frequently as a color-binding error. This covariant relationship was more evident with stronger differential effects of the preceding stimulus, as in the different combination condition and the monocular presentation of the same combination condition. This relationship suggests that the percepts reported as color-binding errors could be described generally as an exclusively dominant test grating with its achromatic region being filled with the color from the suppressed grating. 

Discussion

This study investigated the integration of color and orientation during onset rivalry using equiluminant chromatic gratings in the IDM paradigm. The results indicated that pre-exposure to a stimulus led not only to exclusive dominance, where only one of the rivalrous stimuli was perceived, but also to color-binding errors. These errors were manifested in the perception of bichromatic CW or CCW gratings. The erroneous binding of color and orientation mirrored that observed in sustained binocular rivalry (Hong & Shevell, 2006; Kang & Shevell, 2008; Shevell, St. Clair, & Hong, 2008; Hong & Shevell, 2009; Kang & Shevell, 2011). When exclusive dominance was reported in this study, the effects of the preceding stimulus could be characterized as either eye-based or feature-based modulation, consistent with previous studies (Brascamp et al., 2007; Abe et al., 2011). The preceding stimulus presumably exerted its modulating effects through progressive sensitivity change or adaptation owing to pre-exposure to the same stimulus attribute(s) as one of the rivalrous test stimuli (Brascamp et al., 2007; Abe et al., 2011). 

In addition to establishing that color-binding errors occur in onset rivalry, the present study revealed important characteristics of these errors. The results showed that color-binding errors occurred frequently when the duration of the rivalrous stimuli was brief (30 ms) and when the preceding stimulus shared the same color and/or orientation with either of the rivalrous stimuli. Notably, these errors did not occur selectively with a specific combination of the preceding color and orientation; they occurred under both the same and different combination conditions. Furthermore, the preceding stimulus could be either presented monocularly or binocularly to induce the errors. Finally, a covariant relationship was found between the perception of color-binding errors and that of exclusive dominance, such that the perceived orientation in color-binding errors generally matched that in exclusive dominance. 

The IDM paradigm used in this study enabled a systematic investigation of how stimulus attributes are involved in both the modulating effects of the preceding stimulus and the resolution of onset rivalry. This was achieved without changing the characteristics of the rivalrous stimuli themselves. Consistent with previous findings (Abe et al., 2011), the present study with equiluminant chromatic gratings also revealed that selectivity for feature combinations was manifested differently in the modulating effect of the preceding stimulus and the feature competition for perceptual dominance. When a preceding stimulus modulated the resolution of onset rivalry, its color and orientation affected the dominance of the rivalrous stimuli almost independently. Thus, the modulating effects were observed under the color- and orientation-congruent conditions where only the color or orientation of the preceding stimulus was the same as one of the rivalrous stimuli (Figures 1b and 2b), although the effects were weak in the orientation-congruent condition. Moreover, the dominant percept in the same and different combination conditions could be explained reasonably well by combining color-based suppression and orientation-based facilitation found in the color- and orientation-congruent conditions. The separate modulating effects of the preceding color and orientation are also consistent with the finding in the same combination condition that pre-exposing one of the rivalrous stimuli not only modulated the exclusive dominance, but also produced color-binding errors (Figure 3c). In this case, the preceding stimulus was exactly the same as one of the rivalrous stimuli. Nonetheless, it induced color-binding errors. These findings suggest that the color-based and orientation-based effects did not operate synergistically; rather, the color-based suppression counteracted the orientation-based facilitation. 

In contrast, feature competition in onset rivalry seems to be selective for color–orientation combinations. This selectivity is demonstrated by the absence of misbinding between color and orientation, and the finding that the reported percepts of exclusive dominance and color-binding errors almost always consisted of the color–orientation combination of one of the rivalrous test gratings. In color-binding errors, the bichromatic percept almost always consisted of one of the rivalrous gratings but also bound with the color from the suppressed grating. Furthermore, even when the preceding stimulus differentially modulated the dominance of color and orientation in the color- and orientation-congruent conditions, respectively, the resultant dominant percept was determined in a manner selective for the color–orientation combination. 

Considering the visual mechanisms underlying color-binding errors, these findings suggest that color rivalry was not resolved independently of orientation rivalry. This independent resolution explanation assumes that some properties of the preceding stimulus benefit one orientation of the rivalrous test stimuli, but that the colors of the rivalrous stimuli were not biased by the preceding stimulus, resulting in a percept of a red–green grating. This situation was potentially present in the orientation-congruent condition. However, as mentioned elsewhere in this article, color-binding errors were observed very rarely in that condition. In the same or different combination conditions, the preceding stimulus could benefit the color of one rivalrous grating and the orientation of the other rivalrous grating. However, that did not result in a misbound single-color percept (i.e., a green CCW or red CW grating). This finding also suggests that the independent resolution hypothesis cannot account for the present findings (but see Holmes et al., 2006). 

Another possible mechanism for color-binding errors involves simultaneous color contrast. According to this explanation, the red color in the test grating induced the opposite green color onto the adjacent gray region, resulting in an apparently bichromatic red–green grating. Although previous studies have shown that simultaneous color contrast can be strong when the stimulus duration is short (e.g., 10 ms) (Kaneko & Murakami, 2012), this explanation also seems unlikely. In a preliminary experiment, we asked all four observers to describe the appearance of the rivalrous test stimuli composed of a blue–gray CW and a red–gray CCW grating using the same stimulus sequence as in the same and different combination conditions (luminance and spatial properties of the stimuli were also the same). All observers confirmed color-binding errors composed of a blue–red bichromatic grating. This observation cannot be explained by simultaneous color contrast, consistent with findings in sustained binocular rivalry (Kang & Shevell, 2008; Hong & Shevell, 2009). Moreover, if simultaneous color contrast were the explanation for the color-binding errors, additional mechanisms would need to be invoked to explain why the perception of the contrast-induced bichromatic grating was suppressed in the color- and orientation-congruent conditions, as well as in the control condition. 

A more plausible explanation posits that color-binding errors arise from mechanisms that are closely related to, or even partially overlap with, those responsible for determining exclusive dominance. This explanation aligns with our observation that color-binding errors appear as an exclusively dominant grating whose achromatic regions are filled with the color of the suppressed grating. This explanation suggests that some properties of the preceding stimulus, such as color, orientation, and/or eye of presentation, separately modulate which one of the rivalrous gratings becomes dominant in the same and different combination conditions. Then, occasionally, the color of the suppressed grating escapes suppression and becomes bound with the dominant grating. This mechanism would account for both the occurrence of color-binding errors and their relationship to exclusively dominant gratings. 

To explain how color-binding errors occur, we must hypothesize that color and orientation become decoupled in the suppressed representation, whereas the color–orientation combination is maintained in the dominant percept. This hypothesis is supported by evidence from previous research on interocular suppression. Studies indicate that different features of a single stimulus can be dissociated when the stimulus is rendered invisible owing to interocular suppression (Hong & Blake, 2009; Zadbood, Lee, & Blake, 2011; Moors, Hesselmann, Wagemans, & van Ee, 2017). For instance, Hong and Blake (2009) demonstrated that when a colored stimulus was suppressed from awareness by interocular suppression, the color of the suppressed stimulus could still be perceived; that is, even when the shape of the stimulus was entirely invisible, its color appeared as a diffuse cloud over the suppressing stimulus. This phenomenon suggests that one stimulus feature (e.g., shape) is more susceptible to interocular suppression than another (e.g., color). In this study, pre-exposure to certain stimulus features likely induced differential sensitivity changes or adaptations in different stimulus attributes, resulting in color suppression and only weak orientation facilitation. This differential sensitivity change might have promoted the decoupling of color and orientation in the suppressed representation. 

A related discussion can be found in research on continuous flash suppression, which can be considered a more potent variant of binocular rivalry. Moors et al. (2017) contended that the representation of the interocularly suppressed stimulus “is fractionated rather than integrated and restricted to basic features of the visual input” (see also Sklar, Deouell, & Hassin, 2018 and Moors et al., 2019). This fractionated representation aligns with our hypothesis about the decoupling of color and orientation in the suppressed grating. Furthermore, the fractionated representation of the suppressed stimulus is presumed to reflect relatively early stages of visual processing. Such an assumption is consistent with our findings that a brief test duration is necessary to observe frequent color-binding errors and that a monocular preceding stimulus can induce these errors. Both conditions, that is, a brief test duration and monocular presentation of the preceding stimulus, are commonly linked with eye-based processing (e.g., Wolfe, 1984), suggesting the involvement of early visual processing. 

In summary, we took an exploratory approach to investigate whether and how exposure to a preceding stimulus could lead to misbinding or color-binding errors in color and orientation competitions during onset rivalry. We systematically manipulated various experimental variables and analyzed the appearance of equiluminant rivalrous stimuli reported by a small group of well-experienced observers. Our findings demonstrated that exposure to certain preceding stimuli could induce color-binding errors, which covaried with exclusive dominance when the preceding stimulus had strong modulating effects. However, the specific conditions that cause exclusive dominance to transition into a color-binding error remain elusive. This gap in our understanding underscores the need for further hypothesis-driven research to elucidate the precise mechanisms underlying these phenomena. 

Acknowledgments

Supported by Japan Society for the Promotion of Science KAKENHI Grant Numbers 26285162, 18K18686, and 20H01781. 

Commercial relationships: none. 

Corresponding author: Eiji Kimura. 

Email: [e-mail]. 

Address: Department of Psychology, Graduate School of Humanities, Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba-shi, Chiba 263-8522, Japan.