Serial dependence requires visual awareness: Evidence from continuous flash suppression

Abstract

The visual system often undergoes a relatively stable perception even in a noisy visual environment. This crucial function was reflected in a visual perception phenomenon—serial dependence, in which recent stimulus history systematically biases current visual decisions. Although serial dependence effects have been revealed in numerous studies, few studies examined whether serial dependence would require visual awareness. By using the continuous flash suppression (CFS) technique to render grating stimuli invisible, we investigated whether serial dependence effects could emerge at the unconscious levels. In an orientation adjustment task, subjects viewed a randomly oriented grating and reported their orientation perception via an adjustment response. Subjects performed a series of three type trial pairs. The first two trial pairs, in which subjects were instructed to make a response or no response toward the first trial of the pairs, respectively, were used to measure serial dependence at the conscious levels; the third trial pair, in which the grating stimulus in the first trial of the pair was masked by a CFS stimulus, was used to measure the serial dependence at the unconscious levels. One-back serial dependence effects for the second trial of the pairs were evaluated. We found significant serial dependence effects at the conscious levels, whether absence (Experiment 1) or presence (Experiment 2) of CFS stimuli, but failed to find the effects at the unconscious levels, corroborating the view that serial dependence requires visual awareness.

Introduction

Visual signals from surrounding environments often enter the human eye in an unstable and discontinuous manner due to such factors as occlusion, eye movements, etc., while the visual system generally perceives visual objects as being stable and continuous in a short term. Take white clouds in a clear blue sky as an example. The white clouds could be perceived as an animal such as a dog within a short period of time, even though they are strongly unstable and ever-changing. This perceptual phenomenon is called as serial dependence, whereby the perception of current visual objects is pulled toward recently perceived objects (Ceylan, Herzog, & Pascucci, 2021; Fischer & Whitney, 2014; Kiyonaga, Scimeca, Bliss, & Whitney, 2017; Liberman, Fischer, & Whitney, 2014; Pascucci et al., 2023). The serial dependence effects have been widely observed across a range of visual attributes (for recent reviews, see Manassi, Murai, & Whitney, 2023; Pascucci et al., 2023), including not only low-level stimuli such as grating orientation (e.g., Fischer & Whitney, 2014; Kondo, Murai, & Whitney, 2022) but also high-level stimuli such as facial emotional expression (e.g., Liberman et al., 2014; Mei, Chen, & Dong, 2019). However, the overwhelming majority of previous studies have focused on the serial dependence effects at the conscious levels, and few studies investigated whether the serial dependence effects would occur at the unconscious levels. 

To our knowledge, only three previous studies explored the role of visual awareness during the perception of serial dependence (Fornaciai & Park, 2019; Fornaciai & Park, 2021; Kim, Burr, Cicchini, & Alais, 2020). Fornaciai and Park (2019) first investigated whether “inducer” stimuli (i.e., dot arrays), which had induced the serial dependence effects at the conscious levels in previous studies (e.g., Cicchini., Mikellidou, & Burr, 2017; Fornaciai & Park, 2018), could also induce the effect when the stimuli were removed from visual awareness via visual backward masking technique. The visual backward masking, whereby a brief visual target (generally tens of milliseconds) is immediately followed by a mask, has widely been considered as a paradigm of examining unconscious visual processes (Kim & Blake, 2005; Tsikandilakis et al., 2023; Van Der Ploeg, Brosschot, Versluis, & Verkuil, 2017). Their results showed that the 50-ms “inducer” stimuli failed to induce the serial dependence effect but induced a robust negative aftereffect when the “inducer” stimuli were masked by subsequent 50-ms pattern masks, whereas under a no-mask condition the “inducer” stimuli successfully induced the serial dependence effect. These results were replicated in their further study (Fornaciai & Park, 2021). In addition, using another paradigm of studying unconscious visual processes—binocular rivalry during which two incompatible visual patterns (e.g., a horizontally and vertically oriented grating are presented to the left and right eye, respectively) are perceived in alternate periods of dominance and suppression (Blake & Fox, 1974; Blake & Logothetis, 2002; Zou, He, & Zhang, , 2016), Kim et al. (2020)—found that the perceptually suppressed rivalrous grating stimuli (i.e., at the unconscious levels) also failed to bias perceived orientation of subsequent test gratings, whereas the perceptually dominant rivalrous grating stimuli (i.e., at the conscious levels) induced the classic serial dependence effect. Except for the above-mentioned backward masking and binocular rivalry technique, no other techniques associated with unconscious visual processes were used to investigate whether serial dependence in visual perception would require the involvement of visual awareness. 

Besides the backward masking and binocular rivalry technique, dozens of other psychophysical techniques have been developed to render a visual stimulus invisible (for reviews, Breitmeyer, 2015; Kim & Blake, 2005). Different suppression techniques may induce different levels of unconscious visual processes, demonstrating the existence of a functional hierarchy (Breitmeyer, 2015; Breitmeyer, Koç, Öğmen, & Ziegler, 2008; Izatt, Dubois, Faivre, & Koch, 2014). For example, Breitmeyer et al. (2008) compared the effectiveness of binocular rivalry suppression and metacontrast suppression to examine their relative levels of unconscious processing, suggesting that binocular rivalry suppression appears at an earlier functional level than metacontrast suppression in the visual pathway. Thus, according to the notion of functional hierarchy of unconscious visual processes, previous findings about the perception of serial dependence at the unconscious levels via the backward masking and binocular rivalry technique may not generalize to other suppression techniques. 

In the past decade, continuous flash suppression (CFS) has become the most popular suppression technique in exploring the perception of unconscious visual processes (for recent reviews, Lanfranco, Rabagliati, & Carmel, 2023; Pournaghdali & Schwartz, 2020), and the number of CFS publications has rapidly increased each year (Wang, Alais, Blake, & Han, 2022). The CFS technique, where a rapidly changing high-contrast stimulus pattern (i.e., a Mondrian-like mask) is often presented to the dominant eye and a target stimulus to the nondominant eye (Fang & He, 2005; Tsuchiya & Koch, 2005), has some vital advantages over other suppression techniques. For example, longer suppression duration (Tsuchiya & Koch, 2005) and deeper suppression depth (Tsuchiya, Koch, Gilroy, & Blake, 2006) can be achieved by means of CFS-masking, relative to other masking techniques such as conventional binocular rivalry. In addition, previous studies have shown that CFS and other suppression techniques may tap onto different neural substrates (Blake, Goodman, Tomarken, & Kim, 2019; Dehaene et al., 2001; Fang & He, 2005; Fogelson, Kohler, Miller, Granger, & Tse, 2014). For example, using an individual difference approach, Blake et al. (2019) demonstrated that CFS and binocular rivalry, as two forms of interocular suppression, were not completely controlled by one common neural substrate. It is unclear whether the serial dependence effect would emerge when visual stimuli could be rendered invisible through the CFS technique. 

By combining the CFS technique and the adjustment method, the current study aimed to investigate whether the serial dependence effects would occur at the unconscious levels. Our results showed that the classic serial dependence effects emerged at the conscious levels, whether absence (Experiment 1) or presence (Experiment 2) of CFS stimuli, but did not emerge at the unconscious levels, corroborating previous findings that serial dependence in visual perception requires visual awareness. 

Experiment 1

Methods

Subjects

Twelve naive subjects (eleven females), ranging from 18 to 22 years (mean age = 19.7 ± 1.2 years), participated in Experiment 1. All subjects had normal or corrected-to-normal visual acuity, and provided written informed consent. The study was approved by the School of Psychology Ethics Committee at Guizhou Normal University, and was in accordance with the Declaration of Helsinki (World Medical Association, 2013). 

Stimuli and apparatus

The Gabor patch was used as the experimental stimulus (Figure 1A). The Gabor stimulus, windowed by a 1° SD Gaussian contrast envelope, had 25% Michelson contrast and a spatial frequency of 0.33 cycles per degree (0.33 cpd). The spatial phase of the Gabor stimulus on each trial was randomly selected from the range of 0° to 360°. To remove the Gabor stimulus from visual awareness, we presented a series of high-contrast colored Mondrian-like masks (8.3° × 8.3°) flashing at 10 Hz (i.e., CFS) to subjects’ dominant eye while the Gabor stimulus was presented to the nondominant eye (Figure 1C) (e.g., Lunghi & Pooresmaeili, 2023; Mei, Dong, Dong, & Bao, 2015; Veto, Schütz, & Einhäuser, 2018). We used a mirror stereoscope to implement the presentation of the dichoptic stimuli. 

Figure 1.

 

Sequence of events for the three types of trials in Experiment 1. (A) The response trial. Subjects were required to view the Gabor stimulus and then report the perceived orientation of the Gabor stimulus by adjusting the orientation of a response bar. (B) The no-response trial. Subjects were required to view only the Gabor stimulus, and no response bar appeared. (C) The suppression trial. The Gabor stimulus to the nondominant eye was masked by the CFS stimulus to the dominant eye, and no adjustment responses were required. Note that, for the sake of visualization, the contrasts of the Gabor stimuli were increased, inconsistent with the actual contrasts in the experiments.

The stimuli were presented on a gamma-corrected CRT monitor (1024 × 768 pixels, 85 Hz, 17 inches), and were generated with MATLAB and PsychToolbox-3 (Brainard, 1997; Pelli, 1997). To facilitate binocular fusion, an outer square frame (8.5° × 8.5°) (Figure 1C), consisting of black and white lines, was presented to each eye and remained throughout the experiment. A green fixation dot (0.5° in diameter) was presented in the center of each half of the screen. Subjects were asked to maintain fixation throughout the experiment. They rested their chin on a chin rest to maintain a stable head position during the experiment. Subjects viewed the stimuli at a distance of 57 cm in a dimly lit room. All stimuli were presented on a gray background. 

Dominant eye test

Before the main experiment, the subjects performed a dominant eye test with a breaking continuous flash suppression (b-CFS) paradigm (Lanfranco et al., 2023; Stein, 2019; Stein, Hebart, & Sterzer, 2011). We used the same CFS stimulus as that in the main experiment to assess subjects’ dominant eye. On each trial, the CFS stimulus was randomly presented to either the left or the right eye, while the other eye was simultaneously presented with a black arrow (1.33° × 0.67°) pointing to the left or right direction randomly. The direction and location (left or right eye) of the arrows were counterbalanced across trials. In each trial, the contrast of the CFS stimulus was gradually ramped down from 100% to 0% within a 6000 ms, while the contrast of the black arrow was gradually ramped up from 0% to 25% within a 2000 ms and then remained constant until the end of the trial. Subjects were asked to press the leftward key (“1”) or the rightward key (“3”) as soon as they perceived the direction of the arrow. Subjects were also asked to refrain from blinking. There was a 1000 ms interval between two successive trials. 

Each subject performed 80 trials (2 arrow directions: leftward/rightward × 2 locations: left/right eye × 20 repeats). Notably, 40 trials were included for each eye. Trials with incorrect direction judgments were excluded. The average response time from the remaining trials for each eye was calculated. The eye with a faster average response time was determined as the dominant eye; the other eye as the non-dominant eye. The dominant eye test took about 10 minutes. 

Procedure

There were three trial types: the response trial, the no-response trial, and the suppression trial. In each response trial (Figure 1A), a Gabor stimulus centered 3.9° horizontally away from the central fixation was presented to the non-dominant eye for about 350 ms (30 frames). The Gabor stimulus was presented either to the left or to the right of the fixation dot within each block, with counterbalancing across blocks. After the Gabor stimulus, a noise mask with a same location as the Gabor stimulus was presented to the non-dominant eye for about 471 ms (40 frames) to minimize visual afterimages. The noise mask is comprised of white noise that is smoothed with a 1° SD Gaussian kernel and windowed by a 1° SD Gaussian envelope. After a 471-ms interval, a white response bar (0.61° wide, windowed by a 1° SD Gaussian envelope) located at the same location was presented to the non-dominant eye. The initial orientation of the response bar was randomly selected in each trial. Subjects were instructed to report the perceived orientation of the Gabor stimulus by adjusting the orientation of the response bar. Pressing the “3” or “1” key on the keyboard represented an adjustment in a counterclockwise or clockwise direction, respectively. Subjects submitted their responses by pressing the space bar. The next trial began after a 471-ms intertrial interval. 

The trial structures of the no-response and suppression trial were same that of the response trial except following changes. For these two trial types, the response bar was removed but the fixation dot was remained and the subjects were not allowed to make any response (Figures 1B and 1C). According to the average reaction time of adjusting grating orientations for an adjustment task in our pilot test, we presented the fixation dot for 1800 ms. For the suppression trial, the CFS stimulus was presented to the dominant eye while the Gabor stimulus was presented to the nondominant eye (Figure 1C). To reduce the possibility of accessing visual awareness for the Gabor stimulus, the contrast of the Gabor stimulus was gradually ramped up from 0% to 25% within about 94 ms (eight frames) and then remained at 25% contrast (note that this manipulation was applied to all three trial types). The subjects were asked to press the space bar as soon as they perceived any part of the Gabor stimulus. 

To measure serial dependence at the unconscious levels, we produced trial sequences by using a counterbalanced design of trial pairs (Fischer & Whitney, 2014; Fritsche, Mostert, & de Lange, 2017; Van Der Burg, Rhodes, & Alais, 2019). The first trial of the pairs was randomly chosen from one of the three trial types (i.e., the response trial, the no-response trial, or the suppression trial), and the second trial of the pairs was always the response trial. In other words, there were three types of trial pairs (i.e., three experimental conditions): a response trial versus a response trial, a no-response trial versus a response trial, and a suppression trial versus a response trial. The first two types were designed to measure serial dependence at the conscious levels, and the third to measure serial dependence at the unconscious levels. Basing upon the three types of trial pairs, we can compute one-back serial dependence effects for the second trial of the pairs at the conscious and unconscious levels. For the first trial of each pair, the orientation of the Gabor stimulus was randomly selected in the range of 0° to 180°; for the second trial of each pair, the orientation of the Gabor stimulus was randomly selected in the range of −60° to 60° relative to the orientation of the Gabor stimulus in the first trial of the pair. For example, if the orientation of the Gabor stimulus in the first trial of a pair was 80°, the orientation of the Gabor stimulus in the second trial of the pair was randomly selected from the range of 20° to 140°. Each subject performed 10 blocks, each block including 54 trial pairs (3 types of trial pairs × 18 repeats). Given that there were the two types of trial pair conditions (i.e., a no-response trial vs. a response trial and a suppression trial vs. a response trial) where the subjects were not allowed to make any response for the first trial of the pairs, we added a response trial as an inducer trial at the beginning of each block. Each block took about seven minutes, and the experiment took about 120 minutes in total. 

Data analysis

Serial dependence effects for the three types of trial pairs were separately computed but a same method was used (Fischer & Whitney, 2014; Fritsche & De Lange, 2019; Manassi, Liberman, Kosovicheva, Zhang, & Whitney, 2018). Given that the design of trial pairs described in the Procedure section was used, we computed only one-back serial dependence effects for the three types of trial pairs representing three different experimental conditions. To evaluate sequential biases in adjustment responses in the second trial of the trial pairs against the orientation of Gabor stimuli in the first trial of the trial pairs, we plotted the response errors of the second trial as a function of the orientation differences between the first and second trial of the trial pairs (Figure 2A). The response error of each trial pair was computed by subtracting the orientation of the Gabor stimulus from the reported orientation in the second trial. Positive and negative response errors represented more clockwise and counterclockwise adjustment responses relative to the directions of the Gabor stimuli, respectively (Y-axis in Figure 2A). Likewise, positive and negative orientation differences represented more clockwise and counterclockwise for the Gabor stimuli of the first trial relative to the Gabor stimuli of the second trial of the trial pairs, respectively (X-axis in Figure 2A). Trials in which the response errors exceeded three standard deviations away from the subject's mean error were excluded (on average, 14.25 ± 3.91 trials were excluded for each subject). The breakthrough trials for the suppression condition were also excluded (three trials were excluded for only one subject). 

Figure 2.

 

(A) DoG curve fits to group data for the response trial pair (blue), the no-response trial pair (red), and the suppression trial pair (green) in Experiment 1. X-axis represents the differences between the orientation of the first trial and the second trial of trial pairs; Y-axis represents adjustment errors from the second trial of trial pairs. The thin lines and lightly shaded regions represent average response errors and standard deviation, respectively. (B) Amplitudes of the serial dependence effects for the three trial types. Significant serial dependence effects at the conscious levels (the response and no-response trial pair) but no significant serial dependence effects at the unconscious levels (the suppression trial pair) were found. The error bars indicate the standard deviation of the mean of the bootstrapped distribution. *p < 0.05.

Next, we pooled the response errors from all subjects and fitted the group data with a first derivative of a Gaussian function (DoG) (Ceylan et al., 2021; Collins, 2019; Fischer & Whitney, 2014; Fritsche & De Lange, 2019). The fits were separately conducted for the three types of trial pairs. The DoG is given by  

\begin{eqnarray} y = h + (x + b)awc{e^{ - {{(w(x + b))}^2}}}, \quad \end{eqnarray}

(1)

where y is the response error of the second trial of the trial pairs, x is the orientation difference of the Gabor stimuli between the first and second trial of the trial pairs, a is the amplitude of the DoG curve, w is the width of the DoG curve, h is the height, b is the intercept, and c is the constant \(\sqrt 2 /{e^{ - 0.5}}\). The constant c is chosen to ensure that the parameter a numerically aligns with the peak amplitude of the DoG curve. The amplitude parameter a represents the strength of serial dependence effects, indicating how much the orientation responses to the Gabor stimuli of the second trial of the trial pairs could be biased by the orientations of the Gabor stimuli on the first trial of the trial pairs. A positive a amplitude indicates an attractive perceptual bias (i.e., serial dependence); a negative a amplitude indicates a repulsive perceptual bias (i.e., a negative aftereffect) (Gibson & Radner, 1937; Thompson & Burr, 2009). The parameter w was considered as a free parameter that we constrained to vary within a range of 0.02 to 0.07 (corresponding to the peaks of the DoG curve in the range of 10° to 35° orientation difference). The parameter h is included to allow us to assess general response biases (independent of biases from stimulus history) (Collins, 2019; Collins, 2020; Collins, 2021; Collins, 2022). A positive or negative h represents more responding biases of clockwise or counterclockwise orientations, respectively. In addition, the coefficient of determination (R²) is used to evaluate how well the DoG model captures the data pattern. 

To evaluate the statistical significance of the perceptual bias for each trial type, we used permutation tests as statistical testing (Ceylan et al., 2021; Collins, 2019; Fischer & Whitney, 2014; Fritsche et al., 2017). Specifically, a single permutation was conducted by firstly randomly shuffling the orientation difference of the response errors (i.e., the parameter y) from each subject's original data. Then we fitted the DoG model to the new dataset and obtained the resulting amplitude parameter a. This permutation procedure was repeated 5000 times, and all values of the amplitude parameter a obtained from these permutations formed a null distribution for each trial type. The distribution was compared with the value of the amplitude parameter a estimated from the empirical data to derive a p value. The p value represents the proportion of the null distribution that is greater than or equal to the measured amplitude of serial dependence when the amplitude parameter a of the empirical data is positive, or less than or equal to the measured amplitude of serial dependence when the amplitude parameter a of the empirical data was negative. The significant levels were set at α = 0.05 (two-sided permutation test) for all three trial types. 

Results

The results of the permutation tests showed that the serial dependence effects emerged on the conditions of the response trial (Figures 2A and 2B; a = 2.02°, p = 0.032, R² = 0.49) and the no-response trial (Figures 2A and 2B; a = 2.26°, p = 0.013, R² = 0.57), consistent with previous reports (Cicchini et al. 2017; Fischer & Whitney, 2014; Manassi et al., 2018). In other words, at the conscious processing levels we observed classic significant serial dependence effects. More importantly, however, at the unconscious levels (i.e., the condition of the suppression trial) we failed to observe a statistically significant serial dependence effect, although there was a trend of attractive bias (Figures 2A and 2B; a = 0.72°, p = 0.680, R² = 0.28). These results indicated that the appearance of the visual serial dependence effect required the involvement of visual awareness. 

Experiment 2

In Experiment 1 we failed to find the serial dependence effect under the suppression condition (i.e., the unconscious condition). However, a potential confounding factor was that an additional CFS stimulus under the suppression condition relative to the response and no-response conditions (i.e., the conscious conditions) may reduce or disrupt the serial dependence effect, independent of whether the Gabor stimulus was rendered invisible by the CFS stimulus. Previous studies found that intervening stimuli that did not cause masking (Fornaciai & Park, 2019) or distracting stimuli that directed attention away from inducer stimuli (Fornaciai & Park, 2018) abolished the serial dependence effects of inducer stimuli. Thus in Experiment 2 we aimed to investigate whether the serial dependence effects would still emerge if additional CFS stimuli were involved in the response and no-response conditions in Experiment 1. 

Methods

Subjects

Twelve subjects (seven females), ranging from 19 to 27 years (mean age = 21.6 ± 2.5 years), participated in Experiment 2. One of them had participated in Experiment 1. 

Stimuli, apparatus, dominant eye test, procedure and data analysis

Same stimuli, apparatus, dominant eye test, procedure and data analysis as those in Experiment 1 were used, except that the CFS stimulus was presented to the nondominant eye and the Gabor stimulus was superposed on the CFS stimulus for the first trial of the two types of trial pairs (i.e., a response trial vs. a response trial and a no-response trial vs. a response trial) in Experiment 1 (Figures 3A and 3B). In other words, despite the presentation of the CFS stimuli, the Gabor stimuli were still visible in these two types of trial pairs, and thus the serial dependence effects at the conscious levels were measured. The third type of trial pair (i.e., a suppression trial vs. a response trial; suppression condition), which was used to measure the serial dependence effect at the unconscious levels, was completely same as that in Experiment 1 (Figure 3C). On average, 12.67 ± 6.20 trials were excluded for each subject (a same exclusion rule as that in Experiment 1 was used). No breakthrough trials were found in the suppression condition. 

Figure 3.

 

Sequence of events for the three types of trials in Experiment 2. Same sequence of events as those in Experiment 1, except that the Gabor stimulus was superposed on the CFS stimulus for the response trial (A) or for the no-response trial (B). Sequence of events for the suppression trial (C) was completely same as that in Experiment 1.

Results

Replicating the finding of Experiment 1, we failed to observe a serial dependence effect for the suppression trial (i.e., at the unconscious levels) (Figures 4A and 4B; a = −0.18°, p = 0.944, R² = 0.18). More importantly, although the CFS stimuli were added for the response and non-response conditions (i.e., at the conscious levels), the serial dependence effects still emerged for these two conditions (Figures 4A and 4B; response: a = 2.76°, p < 0.001, R² = 0.70; non-response: a = 1.59°, p = 0.040, R² = 0.68). These findings, to some degree, decreased the possibilities of the CFS stimuli as a potential confounding factor in the suppression condition. The findings of Experiment 2 again corroborated the importance of the involvement of visual awareness in serial dependence. 

Figure 4.

 

(A) DoG curve fits to group data for the response trial pair (blue), the no-response trial pair (red), and the suppression trial pair (green) in Experiment 2. (B) Amplitudes of the serial dependence effects for the three trial types. Although additional CFS stimuli were presented, significant serial dependence effects were still found at the conscious levels (the response and no-response trial pair). Replicating the result of Experiment 1, we failed to find significant serial dependence effects at the unconscious levels (the suppression trial pair). *p < 0.05; ***p < 0.001.

Discussion

By using the CFS technique to render the visual stimuli invisible in the adjustment tasks, the two experiments of the current study failed to find the serial dependence effects at the unconscious levels. Meanwhile, the perceived orientation of the grating stimulus on the current trial was strongly pulled toward the visible orientation of the previous trial (i.e., a significant serial dependence effect at the conscious levels), regardless of whether the adjustment task was performed or not (i.e., on the response or no-response trial). These results provide compelling evidence that serial dependence in visual perception requires visual awareness. 

The null results about serial dependence at the unconscious levels are consistent with those results of previous studies using other suppression techniques, i.e., backward masking (Fornaciai & Park, 2019; Fornaciai & Park, 2021) and binocular rivalry (Kim et al., 2020). Different suppression techniques represent different levels of unconscious processes. Taken together with these studies, we support the notion that serial dependence in visual perception could be absent at different unconscious levels. More future work is needed to confirm the notion because of the existence of dozens of other suppression techniques (Breitmeyer, 2015; Kim & Blake, 2005). 

The findings of our current study that CFS suppressed serial dependence may originate from their partly common neural substrates: primary visual cortex (V1). On the one hand, CFS influences visual awareness by reducing the gain of stimulus-evoked neural responses in human V1 (Yuval-Greenberg & Heeger, 2013); on the other hand, the orientation BOLD signal of human V1 recorded using functional magnetic resonance imaging displayed similar serial dependence effects as behavioral results (St. John-Saaltink, Kok, Lau, & de Lange, 2016). Thus the suppression role of CFS in V1 may result in the null result of serial dependence at the unconscious levels. The other suppression techniques used so far in previous studies about serial dependence (i.e., backward masking and binocular rivalry) also involved the suppression of V1 signals but had different suppression mechanisms (Fornaciai & Park, 2019; Fornaciai & Park, 2021; Kim et al., 2020). Backward masking influences visual awareness by disrupting reentrant processing in V1 from higher visual areas (e.g., Boehler, Schoenfeld, Heinze, & Hopf, 2008; Fahrenfort, Scholte, & Lamme, 2007; Nakashima, Kanazawa, & Yamaguchi, 2021), whereas binocular rivalry does so by initiating interocular competition among monocular neurons in V1 (e.g., Blake, 1989; Carlson et al., 2023; Tong & Engel, 2001). Direct future work is to investigate whether the orientation BOLD signal could display serial dependence effects under a CFS-masking condition. 

Unlike serial dependence, interestingly, visual adaptation as an opposite bias (i.e., repulsive bias or negative aftereffect) reflecting effects of recent perceptual history can successfully escape from the suppression of the CFS masking (Bahrami, Carmel, Walsh, Rees, & Lavie, 2008; Kanai, Tsuchiya, & Verstraten, 2006; Mei et al., 2015). Using the visual backward masking technique, Fornaciai and Park (2019); Fornaciai and Park (2021) demonstrated no serial dependence effects but significant negative aftereffects when the awareness of the stimuli was suppressed. However, in the two experiments of the current study we failed to find significant serial dependence effects or negative aftereffects, consistent with the findings of Kim et al. (2020) in which the binocular rivalry technique was used. These inconsistent findings about negative aftereffects in the research of unconscious serial dependence effects may be likely to originate from different experimental manipulations. For example, the orientations of grating stimuli were relatively random across trials in our study and Kim et al. (2020), whereas the inducer stimuli of inducing serial dependence were relatively stable (e.g., only two kinds of inducer dot arrays were randomly presented across trials) in Fornaciai and Park (2019); Fornaciai and Park (2021). The latter were more likely to result in a gradual accumulation of adaptation effects in the suppression condition, and thus negative aftereffects were observed in their studies. If both serial dependence (attractive bias) and visual adaptation (repulsive bias) operate at the same time in perceptual processing (Fornaciai & Park, 2019; Fritsche, Spaak, & De Lange, 2020; Moon & Kwon, 2022; Rafiei, Hansmann-Roth, Whitney, Kristjánsson, & Chetverikov, 2021; Taubert, Alais, & Burr, 2016), an open question is what determines absence in serial dependence but presence in visual adaptation when visual stimuli could be suppressed from visual awareness. 

We found the significant serial dependence effects under the “no-response” conditions of the two experiments in which the subjects were allowed to make no response toward a previous trial (i.e., the condition better matched the design of the suppression condition). This result replicated previous studies in which “no-response” conditions were also used to examine serial dependence (Fischer & Whitney, 2014; Manassi et al., 2018; Pascucci et al., 2019). A parsimonious interpretation of these results is that the occurrence of serial dependence does not necessarily need an actual overt motor response in the previous trial. The findings that the serial dependence effect can be induced in the “no-response” conditions in previous studies inspired us to combine the CFS paradigm and the adjustment method in exploring serial dependence at the unconscious levels, given that the CFS-masking and the adjustment task are the most widely used tools in the research field of unconscious processing and serial dependence, respectively. Some other explanations may account for the serial dependence effect under the “no-response” condition, such as decision inertia (Pascucci et al., 2019), high-level priors (Cicchini, Benedetto, & Burr, 2021; St. John-Saaltink et al., 2016; Trübutschek & Melloni, 2023) or perceptual templates (Murai & Whitney, 2021). Take decision inertia as an example. Decision inertia is considered as a post-perceptual process in which previous perceptual decisions automatically form an internal tendency guiding current perception (Akaishi, Umeda, Nagase, & Sakai, 2014). Specifically, in the two experiments of our current study, the three experimental conditions (i.e., three trial pairs) included 67% of response trials and 33% of no-response trials, likely resulting in the formation of decisional tendency because of higher probability for response trials. The decision tendency persisted over time and led to the emergence of the serial dependence effect for the no-response condition. Similarly, a significant serial dependence effect was also found under the “no-response” condition in Experiment 2 of Fischer and Whitney (2014), where a high proportion of response trials was also used (75% response trials vs. 25% non-response trials). Interestingly, in Experiment 5 of Pascucci et al. (2019) a significant repulsive effect rather than an attractive serial dependence effect was found for the no-response trials. The probable reason is that a low proportion of responses of stimuli (i.e., in 80% trials of their experiment participants only reported the orientation of the last one of a sequence of six grating stimuli) failed to form implicit decisional inertia. In comparison, the absence of serial dependence effects under the CFS suppression condition would be due to the fact that suppressing conscious awareness of previous stimuli have interfered with the formation of decisional inertia or high-level priors or perceptual templates. More research needs to be conducted to distinguish these explanations. 

Conclusions

In summary, using the CFS-masking technique and the adjustment method, the present study found classic serial dependence effects at the conscious levels but failed to find the serial dependence effect at the unconscious levels. Taken together with previous studies, these results demonstrate that the emergence of the visual serial dependence effects most likely requires visual awareness. 

Acknowledgments

Supported by the National Natural Science Foundation of China (32160198), Guizhou Provincial Science and Technology Foundation (ZK[2021]119), and New Talent Foundation of Guizhou Normal University ([2022]10). The authors declare no conflicts of interest. 

Commercial relationships: none. 

Corresponding author: Gaoxing Mei. 

Email: [e-mail]. 

Address: School of Psychology, Guizhou Normal University, Huaxi University Town, Guian New District, Guiyang 550031, PR China.