Introduction

Although a substantial amount of research suggests that the unit of attentional selection is spatial location, increasing evidence has demonstrated that attentional selection can also be deployed to entire objects (Chen, 2012). The most widely used paradigm to this evidence comes from the double-rectangle cueing paradigm developed by Egly et al. (1994). In this paradigm, two parallel rectangles are presented and participants’ spatial attentions are cued to one end of one of the rectangles. The general finding, i.e., object-based attentional effect, is that the target presented at uncued location within the cued rectangle is detected more rapidly than is physically equivalent target presented within the uncued rectangle. Using this paradigm, this object-based attentional effect has been supported by numerous psychophysical (Chou & Yeh, 2012; Nah et al., 2018; Zhang & Fang, 2012), physiological (Pooresmaeili & Roelfsema, 2014; Roelfsema et al., 1998), and brain imaging (Müller & Kleinschmidt, 2003; Zhang et al., 2017) studies. According to these findings, sensory enhancement hypothesis has proposed that attention directs to one location in an object can spread throughout the whole object, resulting in enhanced quality of the sensory representation of the selected object and more efficient processing of the features that belong to that object (Chen, 2012). This hypothesis is consistent with the predictions of both the incremental grouping (Roelfsema, 2006) and feature integration (Treisman & Gelade, 1980) models.

Complementing the sensory enhancement hypothesis, the attentional prioritization and shifting hypotheses were proposed as the other two main interpretations regarding the mechanism that give rise to object-based attentional effect. The attentional prioritization account emphasizes that the visual search starts from the locations within an already attended object and biases the attentional scanning order (Drummond & Shomstein, 2010; Nah & Shomstein, 2020; Shomstein & Yantis, 2002, 2004), while the attentional shifting account stresses that attentional shifts between objects, relative to within an object, require higher cost (Brown & Denney, 2007; Lamy & Egeth, 2002). Although these hypotheses manifest themselves differently in various experimental operations, such as the spatial uncertainty of targets (Chou & Yeh, 2018; Nah & Shomstein, 2020; Shomstein & Yantis, 2002, 2004) and the requirement of attentional resources (Brown & Denney, 2007), all of them don’t deny the facilitation mechanism induced by within-object condition. However, how the facilitation mechanism works in the within-object condition, or how attention spreads within the object remains unclear.

Accordingly, Roelfsema and colleagues propose that object-based attention spreads within a given object as a gradual manner in the curve-tracing task (Houtkamp & Roelfsema, 2010; Pooresmaeili & Roelfsema, 2014; Roelfsema et al., 1998) and parsing tasks with natural scenes (Korjoukov et al., 2012), 2D images (Jeurissen et al., 2016), and horseshoe-like objects (Ekman et al., 2020). Notably, the perceptual grouping task in all those studies may have involved some degree of endogenous attention since participants needed to track or parse the target curve (image) to determine whether the cue and target belong to the same or different objects. This departs from the original double-rectangle experiments (Egly et al., 1994) which manipulated the attention in an exogenous manner, and calls for experiments to examine such object-based attention spread effect while tease apart the confusions from endogenous attention. In addition, several studies have shown that object-based attention can be generated in lower subcortical region (Strommer et al., 2020) and even guided by an invisible object (Chou & Yeh, 2012; Norman et al., 2013; Zhang & Fang, 2012). It has been widely accepted that rendering a stimulus invisible could maximally (although not completely) reduce various top-down influences (Huang et al., 2020; Wang et al., 2021, 2022; Zhang et al., 2012). Therefore, an important question in this regard is whether and how the exogenous object-based attention spreads without the contamination by top-down signals. To this end, our study here examined whether and how this spread effect interacted with awareness.

In addition, a number of previous studies have indicated that the spread of attention within an object can be guided not only by actual low-level physical boundaries but also by a higher perceptual representation of the object, such as the Kanizsa illusory objects (Martínez et al., 2007; Moore et al., 1998). Since the Kanizsa illusory objects are formed by Kanizsa inducers that produce illusory rather than real object boundaries, perceptual grouping mechanisms are required in order to form a coherent representation of the object (Pak et al., 2020). In other words, the facilitatory influence of attention spreads throughout this Kanizsa-defined object is likely to be controlled at the later or high-level processing (after object forms are integrated) and in this case, whether and if so, how the attentional spread within an illusory object interacts with awareness. Are there different spread processes within an object defined by the real and illusory boundaries?

To address these questions, we used a modified spatial cuing paradigm with the real (Experiment 1) and illusory (Experiment 2) UOs (Fig. 1A) to test the spread of object-based attention and used a backward masking paradigm with low- or high-luminance masks to manipulate the visibility of UOs in both experiments. In our study, the uncued-ends and uncued-middles of UOs, the possible locations of the cue and target, not only locate at an iso-eccentric distance from the fixation but also have an equal distance for the required shift of spatial attention directed by the cue, which thus offered us a unique opportunity to examine whether attention gradually spreads within a given object, i.e., within the UO, attention spreads from its cued-end to uncued-end via its uncued-middle. Results indicated that, despite the visibility (visible or invisible) of UOs, participants in both experiments showed a faster response of the target in the uncued-middle than that in the uncued-end of UOs. Our findings thus indicate an awareness-independent gradual spread of object-based attention, thereby furthering our understanding of the automatic spreading account for object-based attention, as well as the distinction between attention and awareness.

Fig. 1
figure 1

Stimuli and experimental protocol. A Real and illusory U-shaped objects (UOs) used in Experiments 1 (top) and 2 (bottom), respectively. Left: the horizontal UOs; Right: the vertical UOs. On each trial, for Experiment 1, one of the two UOs was equiprobably and randomly present and its symmetrical one was absent (indicated by the dashed line, which was not displayed during the experiments); for Experiment 2, six Kanizsa inducers were always present: three of them (indicated by yellow color here; their color were not changed during the experiments) can be perceived as an illusory UO and the rest of inducers were randomly rotated such that they cannot be perceived as an illusory UO. Both four ends and two middles of UOs, with an iso-eccentric distance from fixation, were the possible location of the cue and target. B The cue and target frames. The cue and the target were the red-hollow and the green-solid circles, respectively. In Experiments 1 and 2, both the cue and target were equiprobably and randomly presented in either the present (the OP condition) or absent (the OA condition) UOs. During each condition, following the cue presented in one of the two ends, the target appeared frequently at the cued-end (CE). It could be also appeared at the uncued-middle (UCM) or uncued-end (UCE) of the UO with equal probability. C Experimental procedure. In both Experiments 1 and 2, each trial began with the fixation cross presented 200 to 1200 ms. Then the cue frame was present for 50 ms, followed by a 50-ms mask (Low- and high-luminance masks rendered the whole UOs visible or invisible to participants, respectively) and another 50-ms fixation interval. Then the target frame was presented for 50 ms, followed by another 50-ms mask. Participants were asked to detect the target as rapidly and correctly as possible and their RTs and accuracies were recorded

Full size image

Methods

Participants

A total of 20 human participants (8 males, 12 females, and 18–24 years old) were involved in the study. All of them participated in Experiments 1 and 2. The sample size was based on those used in previous studies on object-based attention in our lab (Zhang et al., 2017; Zhang & Fang, 2012) and a priori power calculation using G*Power program (Faul et al., 2009). The power analysis indicated that a sample of 20 in our experiments would be sufficient to detect a medium-size effect (f = 0.25) in a within-participants analysis of ANOVA with the power of 0.8. They had normal or corrected-to-normal visual acuity and were naive to the purpose of the study. They gave written, informed consent, and our procedures and protocols were approved by the human participants review committee of School of Psychology at South China Normal University.

Stimuli

The two symmetrical U-shaped objects (UOs, subtended 15.684° ×13.750° in visual angle) defined by the real and illusory boundaries were used in Experiments 1 and 2, respectively (Fig. 1A). On each trial, for Experiment 1, one of the two UOs was randomly present and its symmetrical one was absent; for Experiment 2, six Kanizsa inducers (subtended 2.188° × 2.188° in visual angle) were always present: three of them can be perceived as an illusory UO (parallel to the present UO in Experiment 1) and the rest of three inducers were randomly rotated from 30° to 330° such that they cannot be perceived as an illusory UO (parallel to the absent UO in Experiment 1). The four ends and two middles of the two UOs, with an iso-eccentric distance from fixation (eccentricity: 12.50° in visual angle), were the possible location of the cue (the red hollow circle, 1.25°× 1.25°) and target (the green solid circle, 1.25°× 1.25°). In Experiments 1 and 2 (Fig. 1B), both the cue and target were equiprobably and randomly presented in either the present (object present condition, OP) or absent (object absent condition, OA) UOs. Note that, on each trial, the cue and target never appeared in the present and absent UOs, respectively (and vice versa, they also never appeared in the absent and present UOs, respectively). Low- and high-luminance masks (subtended 28.125° ×28.125° in visual angle), which were Fourier transform images, rendered the whole UOs visible or invisible (confirmed by a two-alternative forced choice [2AFC] test before each experiment, see Supplemental Methods) to participants, respectively (Fig. 1C).

Procedure

Visual stimuli were displayed on an IIYAMA color graphic monitor (model: HM204DT; refresh rate: 60 Hz; resolution: 1,280 × 1,024; size: 22 inches) at a viewing distance of 57 cm. Participants’ head position was stabilized using a chin rest. A white fixation cross (0.625° × 0.625°) was always presented at the center of the screen throughout the experiment.

The current study consisted of Experiments 1 and 2, with the same procedure except for the stimuli. Experiments 1 and 2 examined the spread of object-based attention using the real and illusory UOs, respectively. Participants participated in Experiments 1 and 2 on two different days, and the order of the two experiments was counterbalanced across participants. In both Experiments 1 and 2, each trial began with the fixation cross presented 200 to 1200 ms. Then the cue frame was present for 50 ms, followed by a 50-ms mask (Low- and high-luminance masks rendered the whole UOs visible or invisible to participants, respectively) and another 50-ms fixation interval. Then the target frame was presented for 50 ms, followed by another 50-ms mask. Participants were asked to detect the target as rapidly and correctly as possible (Fig. 1C) and their reaction times (RTs) and accuracies were recorded.

Both Experiments 1 and 2 consisted of four sessions: the visible-horizontal UOs, invisible-horizontal UOs, visible-vertical UOs, and the invisible-vertical UOs; the order of these four sessions was counterbalanced across participants. Each session consisted of 10 blocks of 144 trials: 72 trials for the OP condition and 72 trials for the OA condition. In each block of each condition, the cue was equiprobably and randomly presented in the middle (24 trials) and two ends (48 trials, 24 trials for each one) of the UO. Following the cue presented in one of the two ends, the target could appear at the cued-end (CE) in 18 trials, uncued-middle (UCM) in 2 trials, uncued-end (UCE) in 2 trials, or be absent in 2 catch trials (any response on a catch trial was recorded as an error) (Fig. 1B). Following the cue presented in the middle of the UO, the target could appear at the cued-middle in 18 trials, two uncued-ends in 4 trials, or be absent in 2 catch trials (note that there was no UCM condition, therefore, this situation was excluded for the analysis). All trials were randomized in a block and participants received the percentage of correct responses after each block.

Results

Before each experiment, we checked the effectiveness of the awareness manipulation by a 2-AFC test. The stimuli and procedure in the 2-AFC test were the same as those in Experiments 1 and 2, except that, participants were asked to judge the opening orientation of the present UOs: leftward or rightward for the horizontal-UOs block; upward or downward for the vertical-UOs block (Fig. 1A). Results showed that, in the invisible condition, participants’ performances were not statistically different from chance (mean percent correct ± standard error of the mean [SEM], Experiment 1: 49.232 ± 0.597%, t(19) = -1.286, p = 0.214, Cohen’s d = 0.288; Experiment 2: 50.195 ± 0.326%, t(19) = 0.600, p = 0.556, Cohen’s d = 0.134), confirming that the UOs were invisible; in the visible condition, by contrast, their performances were more accurate (Experiment 1: 98.151 ± 0.384%; Experiment 2: 95.178 ± 0.676%), indicating that our awareness manipulation was effective for both visible and invisible conditions.

Experiment 1: real objects

The percentage of correct responses to target was 98.441% in visible and 98.172% in invisible conditions. There was no difference in accuracy across conditions. Correct RTs shorter than 150 ms, longer than 1000 ms and beyond three standard deviations from the mean RT in each condition were removed. The removal rates were 3.010% and 3.219% in visible and invisible conditions respectively. To quantitatively examine the relationship between awareness and the spreading pattern of object-based attention, participants’ mean RTs (see Table 1) were submitted to a three-way repeated-measures ANOVA with the awareness (visible vs. invisible), object appearance (OP vs. OA), and the target location (CE vs. UCM vs. UCE) as within-participants factors. The interaction of these three factors was not significant (F(2,38) = 0.933, p = 0.395, ηp2 = 0.047). Therefore, for both visible and invisible conditions, participants’ mean RTs were submitted to a two-way repeated-measures ANOVA with the object appearance (OP vs. OA) and the target location (CE vs. UCM vs. UCE) as within-participants factors. Differences between conditions were further analyzed with planned pairwise comparisons. Greenhouse-Geisser correction was performed when necessary.

Table 1 RTs (mean ± standard error of the mean [SEM]) of each condition in Experiment 1
Full size table

For the visible condition (Fig. 2A, left), results showed that the main effect of object appearance (F(1,19) = 6.290, p = 0.021, ηp2 = 0.249 ), the main effect of target location (F(2,38) = 18.317, p < 0.001, ηp2 = 0.491), and the interaction between these two factors (F(2,38) = 14.969, p < 0.001, ηp2 = 0.441) were all significant. Post hoc paired t tests showed that the RTs in the UCM and UCE trials were significantly greater than those in the CE trials for both OP (UCM vs. CE: t(19) = 2.594, p = 0.018, Cohen’s d = 0.580; UCE vs. CE: t(19) = 4.991, p < 0.001, Cohen’s d = 1.116) and OA (UCM vs. CE: t(19) = 5.726, p < 0.001, Cohen’s d = 1.280; UCE vs. CE: t(19) = 3.193, p = 0.005, Cohen’s d = 0.714) conditions, indicating the classical spatial attention effect. The RT in the UCE trials was significantly larger than that in the UCM trials for OP condition (t(19) = 4.027, p = 0.001, Cohen’s d = 0.900), but smaller for OA condition (t(19) = -2.864, p = 0.010, Cohen’s d = 0.640). For the invisible condition (Fig. 2A, right), results showed that the main effect of object appearance was not significant (F(1,19) = 0.191, p = 0.667, ηp2 = 0.010), but the main effect of target location (F(2,38) = 21.592, p < 0.001, ηp2 = 0.532) and the interaction between these two factors (F(2,38) = 10.677, p = 0.001, ηp2 = 0.360) were both significant. Post hoc paired t tests showed that the RTs in the UCM and UCE trials were significantly greater than those in the CE trials for both OP (UCM vs. CE: t(19) = 3.054, p = 0.007, Cohen’s d = 0.683; UCE vs. CE: t(19) = 7.358, p < 0.001, Cohen’s d = 1.645) and OA (UCM vs. CE: t(19) = 4.231, p < 0.001, Cohen’s d = 0.946; UCE vs. CE: t(19) = 4.567, p < 0.001, Cohen’s d = 1.021) conditions, also indicating the classical spatial attention effect. The RT in the UCE trials was significantly larger than that in the UCM trials for OP condition (t(19) = 4.254, p < 0.001, Cohen’s d = 0.951), but smaller for OA condition (t(19) = -2.135, p = 0.046, Cohen’s d = 0.477). Together, these results suggested that attention could gradually spread within the real object and this gradual spread was independent of awareness.

Fig. 2
figure 2

Results of Experiment 1: Real Objects. A Mean RTs in Experiment 1 are shown for CE, UCM, and UCE in the OP and OA trials for visible (left) and invisible conditions (right). B The ΔRT quantified as the RT difference between UCE and UCM in the OP and OA trials for visible and invisible conditions. C The gradual spread effect (GSE = ΔRTOP – ΔRTOA) for visible and invisible conditions. Error bars denote 1 SEM calculated across participants

Full size image

Subsequently, to directly quantify the gradual spread of object-based attention, we analyzed the RT difference (ΔRT) between UCE and UCM trials in both the OP and OA conditions (Fig. 2B). The ΔRT was calculated as follows: ΔRT = RTUCE – RTUCM, where RTUCE and RTUCM are the mean RTs for the UCE and UCM trials, respectively. We hypothesized that, for the OP condition, if attention gradually spreads within the object, participants’ detection of the target in the UCM trials should be faster than that in the UCE trials. The ΔRT then should be significantly higher than zero. However, if attention doesn’t gradually spread within the object, the ΔRT should not be significantly different than zero. For the OA condition, the ΔRT should not be significantly different than zero since no object was presented in this condition. Results showed that, for the OP condition, ΔRTs were significantly higher than zero in both visible (t(19) = 4.027, p = 0.001, Cohen’s d = 0.900) and invisible (t(19) = 4.254, p < 0.001, Cohen’s d = 0.951) conditions, further supporting an awareness-independent gradual spread of object-based attention. Surprisingly, the OA condition demonstrated a revised pattern to the OP condition by showing that ΔRTs were significantly lower than zero in both visible (t(19) = -2.864, p = 0.010, Cohen’s d = 0.640) and invisible (t(19) = -2.135, p = 0.046, Cohen’s d = 0.477) conditions (Fig. 2B). In other words, participants’ detection of the target in the UCE trials was significantly faster than that in the UCM trials, suggesting a potential difference in spatial attention between the UCE and UCM trials. Although this difference in spatial attention couldn’t account for our awareness-independent gradual spread of object-based attention, to further exclude this issue, we computed the gradual spread effect (GSE) to quantify how much the ΔRT showed in the OP condition relative to that in the OA condition in both visible and invisible conditions (Fig. 2C). The GSE was calculated as follows: GSE = ΔRTOP – ΔRTOA, where ΔRTOP and ΔRTOA are the ΔRT for the OP and OA conditions, respectively. Results showed that GSEs were significantly higher than zero in both visible (t(19) = 5.018, p < 0.001, Cohen’s d = 1.122) and invisible (t(19) = 3.815, p = 0.001, Cohen’s d = 0.853) conditions, and more importantly, there was no significant difference in GSE between these two conditions (t(19) = 1.542, p = 0.140, Cohen’s d = 0.345). These results further confirmed an awareness-independent gradual spread of object-based attention within the real objects.

Experiment 2: illusory objects

The percentage of correct responses to target was 98.483% in visible and 98.295% in invisible condition. There was no difference in accuracy across conditions. Correct RTs shorter than 150 ms, longer than 1000 ms and beyond three standard deviations from the mean RT in each condition were removed. The removal rates were 1.920% and 2.017% in visible and invisible conditions, respectively. Table 2 shows the RTs (mean RTs ± standard error of the mean [SEM]) of each condition.

Table 2 RTs (mean ± standard error of the mean [SEM]) of each condition in Experiment 2
Full size table

Similar to Experiment 1, first, the same three-way repeated-measures ANOVA showed that the interaction of the three factors was not significant (F(2,38) = 1.881, p = 0.172, ηp2 = 0.090). Therefore, for both visible and invisible conditions, participants’ mean RTs were submitted to the same two-way repeated-measures ANOVA analyses. Second, for the visible condition (Fig. 3A, left), results showed that the main effect of object appearance was not significant (F(1,19) = 0.111, p = 0.742, ηp2 = 0.006 ), but the main effect of target location (F(2,38) = 68.703, p < 0.001, ηp2 = 0.783) and the interaction between the two factors (F(2,38) = 28.833, p < 0.001, ηp2 = 0.603) were both significant. Post hoc paired t tests showed that the RTs in the UCM and UCE trials were significantly greater than those in the CE trials for both OP (UCM vs. CE: t(19) = 7.566, p < 0.001, Cohen’s d = 1.701; UCE vs. CE: t(19) = 11.191, p < 0.001, Cohen’s d = 2.490) and OA (UCM vs. CE: t(19) = 7.417, p < 0.001, Cohen’s d = 1.655; UCE vs. CE: t(19) = 6.583, p < 0.001, Cohen’s d = 1.470) conditions, indicating the classical spatial attention effect. The RT in the UCE trials was significantly larger than that in the UCM trials for OP condition (t(19) = 6.509, p < 0.001, Cohen’s d = 1.448), but smaller for OA condition (t(19) = -3.716, p = 0.001, Cohen’s d = 0.823). For the invisible condition (Fig. 3A, right), results showed that the main effect of object appearance was not significant (F(1,19) = 1.444, p = 0.244, ηp2 = 0.071), but the main effect of target location (F(2,38) = 48.972, p < 0.001, ηp2 = 0.720) and the interaction between these two factors (F(2,38) = 19.425, p < 0.001, ηp2 = 0.506) were both significant. Post hoc paired t tests showed that the RTs in the UCM and UCE trials were significantly greater than those in the CE trials for both OP (UCM vs. CE: t(19) = 4.541, p < 0.001, Cohen’s d = 1.007; UCE vs. CE: t(19) = 8.773, p < 0.001, Cohen’s d = 1.969) and OA (UCM vs. CE: t(19) = 7.122, p < 0.001, Cohen’s d = 1.613; UCE vs. CE: t(19) = 6.433, p < 0.001, Cohen’s d = 1.478) conditions, also indicating the classical spatial attention effect. The RT in the UCE trials was significantly larger than that in the UCM trials for OP condition (t(19) = 4.575, p < 0.001, Cohen’s d = 1.019), but not for OA condition (t(19) = -1.654, p = 0.115, Cohen’s d = 0.361). These results suggested that attention could gradually spread within the illusory object and this gradual spread was independent of awareness. Third, for the OP condition, ΔRTs were significantly higher than zero in both the visible (t(19) = 6.509, p < 0.001, Cohen’s d = 1.455) and invisible(t(19) = 4.575, p < 0.001, Cohen’s d = 1.023) conditions, further supporting an awareness-independent gradual spread of object-based attention. For the OA condition, ΔRTs were significantly lower than zero in visible condition (t(19) = -3.716, p = 0.001, Cohen’s d = 0.831) but not in invisible condition (t(19) = -1.654, p = 0.115, Cohen’s d = 0.370) (Fig. 3B). Finally, the GSEs were significantly higher than zero in both visible (t(19) = 6.894, p < 0.001, Cohen’s d = 1.541) and invisible (t(19) = 4.874, p < 0.001, Cohen’s d = 1.090) conditions, and more importantly, there was also no significant difference in GSE between the two conditions (t(19) = 1.751, p = 0.096, Cohen’s d = 0.391) (Fig. 3C). These results further confirmed an awareness-independent gradual spread of object-based attention within the illusory objects.

Fig. 3
figure 3

Results of Experiment 2: Illusory Objects. A Mean RTs in Experiment 2 are shown for CE, UCM, and UCE in the OP and OA trials for visible (left) and invisible conditions (right). B The ΔRT quantified as the RT difference between UCE and UCM in the OP and OA trials for visible and invisible conditions. C The gradual spread effect (GSE = ΔRTOP – ΔRTOA) for visible and invisible conditions. Error bars denote 1 SEM calculated across participants

Full size image

Experiment 1 versus experiment 2

To further quantitatively examine the effect of “visual objectness” (real UO vs. illusory UO) in the gradual spreading of object-based attention, the GSEs within the real (Experiment 1) and illusory (Experiment 2) objects were submitted to a two-way repeated-measures ANOVA with the awareness (visible vs. invisible) and the “visual objectness” (real UO vs. illusory UO) as within-participants factors. Results showed that the main effect of awareness was significant (F(1,19) = 5.541, p = 0.029, ηp2 = 0.226), but the main effect of “visual objectness” (F(1,19) = 0.031, p = 0.861, ηp2 = 0.002) and the interaction between the two factors (F(1,19) = 0.083, p = 0.776, ηp2 = 0.004) were not significant. Therefore, our results of two experiments not only support a gradual spread of object-based attention but also reveal that this gradual spread is independent of “visual objectness” (whether the object is defined as the real or illusory boundaries, Lengyel et al., 2021) and conscious access to objects.

Discussion

The present results provide support for a gradual spread of attention within a given object defined by both the real and illusory boundaries, and further reveal that this gradual spread is independent of awareness. Our gradual spread of object-based attention evident here was indexed by a faster response of the target in the uncued-middle than uncued-end of UOs (Figs. 2 and 3). Importantly, this faster effect cannot be explained by the spatial attention. First, the uncued-middle and uncued-end not only locate at an iso-eccentric distance from the fixation but also have an equal distance for the required shift of spatial attention directed by the cue (Fig. 1A). Second, compared to the OP condition, there was a non-significant or reversed difference in participants’ detection of the target between the uncued-middle and uncued-end during the OA condition (Figs. 2 and 3). Although speculative, it is plausible that this reversed difference in the OA condition may result from a benefit that the cued-end and uncued-end were at symmetrical locations from the fixation (note that the cued-end and uncued-middle were at asymmetrical locations from the fixation). Following the cue, the symmetrical uncued-end in visual space may attract more cognitive resource or induce faster attentional shift than the asymmetrical uncued-middle (Kootstra et al., 2008; Palmer & Hemenway, 1978), and thus results in a faster response of the target in the uncued-end than that in the uncued-middle. Critically, it is important to note that this symmetrical benefit also existed during the OP condition and we should have observed the same results. However, our data showed that this was not the case. Thus, the reversed difference in the OA condition further strengthened our conclusion that the spread of attention within an object was a gradual processing. Finally, participants’ response to the target in the uncued-middle was shorter than that appearing in the uncued-end, even after subtracting their respective baselines (i.e., participants’ reaction times during the OA condition, Figs. 2 and 3).

In addition, a number of previous studies have suggested that shifts in attention may depend on the hemifield of the target (Cavanagh & Alvarez, 2005; Shipp, 2011), therefore, one might argue that our observed gradual spread of object-based attention could be derived by the hemifield difference between the uncued-middle and uncued-end targets, relative to the cue. In our study, for the horizontal UOs, the uncued-middle target and the cue were presented across the left-right hemisfield but within the same upper/lower hemisfield whereas the uncued-end target and the cue were presented within the same left/right hemisfield but across the upper-lower hemisfield; for the vertical UOs, notably, the situation was reversed (Fig. 1A). If the hemisfield difference explanation could account for our results, then we should have observed different/reversed spread effects between the horizontal and vertical UOs. However, our results argue against this explanation by showing the same qualitative conclusion between the two (Figs. S1 & S2).

The most parsimonious account of our results is that the spread of attention within an object was a gradual processing even when the object is not consciously perceived. Our results not only support the sensory enhancement hypothesis, which proposes that attention directs to one location in an object can spread throughout the whole object, resulting in enhanced quality of the sensory representation of the selected object and more efficient processing of the features that belong to that object (Chen, 2012), but also further indicate that this enhanced sensory representation spreads within a given object as a gradual manner. In addition, one should note that our results cannot directly support the other two main interpretations regarding to the object-based attentional effect: attentional prioritization and shifting hypotheses. The former emphasizes that the visual search starts from the locations within an already attended object and the uncertainty of targets can change the assignment of attention (Shomstein & Yantis, 2002, 2004), while the latter stresses that attentional shifts between objects, relative to within an object, require higher cost (Brown & Denney, 2007; Lamy & Egeth, 2002). On the one hand, although our experiments set a high and low probabilities in the cued and un-cued conditions and indicated a spatial attention benefit within an object, we did not set a different probability between the uncued-middle and uncued-end conditions (Shomstein & Yantis, 2004) to change the assignment of object-based attention. One the other hand, our study didn’t manipulate the appearance of pre-cues or the number of targets to explore the attentional resources’ distribution problem. Thus, in the future, it is worthwhile to examine how the object-based attention spreads within objects under the consideration of attentional priority and shifting hypotheses.

Our findings not only are consistent with previous studies using the perceptual grouping task, such as the curve-tracing (Houtkamp & Roelfsema, 2010; Pooresmaeili & Roelfsema, 2014; Roelfsema et al., 1998) and parsing (Ekman et al., 2020; Jeurissen et al., 2016; Korjoukov et al., 2012) tasks by showing a gradual spread of attention over the relevant object so that all its features can be bound into a unified percept of the object, but also support the incremental grouping (Roelfsema, 2006) model, which proposes that our visual system establishes the representation of an object by grouping its attended-features/parts to unattended-features/parts incrementally with attention. In other words, attention directed to one location in an object gradually spreads throughout the entire object that is caused by the transitivity of a serial, incremental grouping process. For example, in our study, the cued-end, uncued-middle, and uncued-end are features/parts of an UO and to establish the incremental groupings, the visual system evaluates the local groupings one at a time, by spreading of object-based attention, first from cued-end to uncued-middle and then from uncued-middle to uncued-end until the entire UO has been labeled with enhanced activity. This time-consuming spread leads to the explicit representation of the UO via recurrent connections between neural populations that belong to the attended UO. Here our results extend this model by showing an awareness-independent gradual spread of object-based attention. In other words, object awareness is not necessary for the spread of object-based attention, and visible and invisible objects can trigger the same gradual spread process. Given rendering a object invisible could maximally (although not completely) reduce various top-down influences (Huang et al., 2020; Wang et al., 2021, 2022; Zhang et al., 2012), our results, therefore, also can be viewed as identifying an automatic spread of object-based attention, supporting previous psychophysical (Chou & Yeh, 2012; Norman et al., 2013; Zhang & Fang, 2012), neurophysiological (Wannig et al., 2011), and brain imaging (Ekman et al., 2020) studies.

Our results also indicated that this automatic and gradual spread of object-based attention was independent of whether the object is defined as the real or illusory boundaries. These findings not only converge with previous neurophysiological (Peterhans & von der Heydt, 1989) and brain imaging (Maertens & Pollmann, 2005) studies indicating that some neurons in early visual cortex respond to illusory contours similarly to real lines, but also support other studies reporting that the illusory contour processing appears to be obligatory (Moore et al., 1998): it could be formed preattentively (Davis & Driver, 1994), occur quickly (Rauschenberger & Yantis, 2001), doesn’t depend on feedback from parietal cortex (Mattingley et al., 1997), and is encoded at early visual processing stages (Senkowski et al., 2005). However, previous imaging (Murray et al., 2002; Stanley & Rubin, 2003) and lesion (De Weerd et al., 1996; Huxlin et al., 2000) studies have conversely reported an involvement of higher-tier visual areas in illusory contour representation, in disagreement with our awareness-independent spread within the object defined by illusory boundaries. One possible explanation is that the backward masking in our study substantially decreased the object related activity in these areas to a level below the threshold of awareness rather than object-based attention. Alternatively, the feedback from these areas, disrupted by the backward masking, might be not necessary for the spread of object-based attention. In addition, the setting of elements was entangled with the perception of illusory contours in our experiments, and how partial elements (but not the illusory contours) of our setting affected the object-based attentional effect was unknown. Several brain imaging studies have focused on the relationship between the illusory object and Kanisza inducer (element), and found the reverse activation of illusory object and partial elements in V1 (Kok & de Lange, 2014; Kok et al., 2016). Thus, our study cannot deny a potential contribution from Kanisza inducers in attention spreads within an object defined by the illusory boundaries and further studies will shed light on these issues using neurophysiological or brain imaging techniques.

In addition, our study also provides evidence for a long-standing debate whether attention and awareness are independent (Koch & Tsuchiya, 2012). Numerous studies have demonstrated that spatial (Huang et al., 2020; Jiang et al., 2006; Wang et al., 2021, 2022; Zhang et al., 2012) and feature-based (Kanai et al., 2006; Melcher et al., 2005) attention is independent of awareness, object-based attention and awareness, however, are often considered to be closely interwoven; attending to an object can visually aware of it and being aware of an object may lead to attention directed toward it. Intriguingly, the present study provides new evidence indicating that object-based attention and awareness are two dissociated functions in the visual system by demonstrating an awareness-independent gradual spread of object-based attention. Based on these findings, in conjunction with existing psychophysical studies (Chou & Yeh, 2012; Norman et al., 2013; Zhang & Fang, 2012), we speculate that the gradual spread of object-based attention without awareness might be intrinsically related to high-level unconscious processing, such as the word meaning (Naccache et al., 2002), scene gist (Li et al., 2002), object category (Almeida et al., 2008), and face emotion (Yang et al., 2007). These high-level unconscious processes in our visual system indicate that object representations seem to have an ecological function—some potentially important objects may be able to guide attention and undergo deeper processing before they enter consciousness.

Our findings can be viewed as identifying an awareness-independent gradual spread of object-based attention. Note that this conclusion is based on the isolated object without directly competition with other objects, especially for our real objects in Experiment 1. Although our results are consistent with several studies showing the same object-based attentional effect using the isolated object also (Baldauf & Desimone, 2014; Barnas & Greenberg, 2016; Collegio et al., 2019; Molholm et al., 2007; Soto & Blanco, 2004), the biased competition model (Desimone & Duncan, 1995) proposes that cortical representations of items within the scene are mutually competed; thus, the strength of a representation is weaker when an object appears in isolation (without competition). The classical object-based attentional effect emerges since an attentional cue that summons spatial attention to a particular object serves as a biasing signal, guides attention, and strengthens the sensory representation at the attended object, allowing it to compete more effectively with other unattended representations. Indeed, using multiple objects presented simultaneously, many previous psychophysical (Chou & Yeh, 2012, 2018; Jeurissen et al., 2016; Nah et al., 2018; Zhang & Fang, 2012), physiological (Ekman et al., 2020; Pooresmaeili & Roelfsema, 2014; Roelfsema et al., 1998), and brain imaging (Müller & Kleinschmidt, 2003; Zhang et al., 2017) studies have supported the object-based attentional effect by showing the enhanced quality of the sensory representation of the selected object and more effectively spreads to other features that belong to that object. Future work is thus needed using the objects both with and without competition to parse how the object-based attention spreads within an object and how this spread interacts with awareness.

In sum, our study provides, to the best of our knowledge, the first experimental evidence supporting an awareness-independent gradual spread of object-based attention, thereby furthering our understanding of the automatic spreading account for object-based attention, as well as the distinction between attention and awareness.