PLoS Biology
Public Library of Science
image
The human visual system differentially represents subjectively and objectively invisible stimuli
DOI 10.1371/journal.pbio.3001241 , Volume: 19 , Issue: 5 , Pages: 0-0
Article Type: research-article, Article History
Abstract

This study shows that the extent of unconscious information processing in human visual cortex is determined by the measurement approach; only subjectively, but not objectively, invisible stimuli are represented at a categorical level.

Stein, Kaiser, Fahrenfort, van Gaal, and Malach: The human visual system differentially represents subjectively and objectively invisible stimuli

Introduction

Determining the function and neural correlates of human consciousness is one of the most challenging topics in psychology and cognitive neuroscience today [13]. The scientific study of consciousness requires pitting conscious processes against comparable unconscious processes [4]. One powerful approach is to compare neural processing between stimuli presented outside conscious awareness and stimuli that are consciously perceived. Although all major theories of consciousness are based on the notion that stimuli can be processed unconsciously [58], the scope and extent of unconscious processing are highly debated, with estimates ranging from low-level perceptual analysis [9,10] to high-level object categorization [1113] and full-blown unconscious cognition and reasoning [14,15]. One important cause of this controversy relates to theoretical disagreements about how to measure consciousness and how to demonstrate absence of conscious awareness [11,16].

The most intuitive approach is to simply ask participants to introspectively report their experience of a barely perceivable (e.g., masked) stimulus [17]. Recent studies adopting such subjective awareness measures found that visual stimuli reported as “invisible” still undergo high-level processing. For example, subjectively invisible stimuli are encoded [18,19] and stored [2022] in visual cortex, enabling above-chance perceptual discrimination even after a memory delay [23,24]. By linking behavioral performance for subjectively invisible stimuli to brain activity, the neural correlates of such “blindsight”-like performance can be recorded in normal observers [25]. Subjective measures assume that an observer’s decision to report awareness accurately distinguishes the presence versus absence of conscious perceptual information. However, this assumption is at odds with research on perceptual decision-making, which shows that subjective measures do not accurately reflect perception, because they are susceptible to decision biases [26,27]. Participants often have a conservative bias [28], such that an “invisible” response merely indicates that a stimulus was relatively difficult to see [29]. Thus, subjective measures may misclassify (partially) conscious stimuli as unconscious. Studies adopting purely subjective measures therefore most likely overestimate the extent of unconscious processing [26,30,31].

To convincingly rule out conscious perception, objective awareness measures based on performance are required [26]. Crucially, these measures should target the key stimulus characteristic of interest [32]. For example, when contrasting neural responses evoked by masked faces and houses, participant’s performance in discriminating the two stimulus categories should not exceed what is expected by chance [33]. However, this approach is also riddled with challenges. First, statistically, scientists have taken a failure to reject the null hypothesis (of chance performance, p > 0.05) as support for the null hypothesis, which is invalid in the standard frequentist hypothesis testing framework. A nonsignificant effect can be related to measurement noise and lack of statistical power rather than genuine invisibility (chance performance). In neuroimaging studies in particular, this represents a serious concern as sample sizes tend to be very small (e.g., only four [34], five [33,35], six [36], seven [37], or eight [38] participants). Further, objective awareness is often only measured “offline” in a block separate from the functional magnetic resonance imaging (fMRI) recordings (or even outside the MRI scanner) [3542], and with considerably fewer trials [35,37,3941] and sometimes also fewer participants [38,40]. Given these issues, it is unknown how stimuli that are genuinely invisible according to the objective definition are represented in different cortical areas.

Here, our first goal was thus to measure the extent of unconscious processing of objectively invisible stimuli in human visual cortex. To ensure sufficient statistical power, we included data from 43 human observers who completed a large number of trials while measuring brain activity evoked by masked faces or houses using a multiband fMRI sequence with fast acquisition time and high spatial resolution. An “online” (trial-by-trial) measure of perceptual discriminability of faces and houses served as an exhaustive [32,43] measure of objective awareness, leading to equal numbers of trials to calculate objective discrimination performance and to evaluate the extent of neural processing of masked face/house stimuli. Perceptual sensitivity was analyzed with Bayesian statistics to establish genuine objective invisibility.

Compared to subjective invisibility, objective invisibility requires strong reduction of stimulus strength, e.g., by brief presentation times, low contrasts, or strong masking, which reduces neural responses. Estimates of unconscious processing, both in behavior and in neural recordings, may thus critically depend on selection and implementation of method and statistics for establishing absence of conscious awareness. However, as objective and subjective awareness measures have not been compared in the same study, it is unclear whether the 2 approaches are merely associated with quantitative differences (e.g., subjective measures allowing for stronger neural responses to unconscious stimuli) or whether important qualitative differences exist (e.g., subjective measures allowing for unconscious processing of distinct higher-level stimulus properties).

Hence, our second goal was to directly compare the two approaches. We used fMRI to measure brain activity evoked by masked faces or houses that were visible, subjectively invisible, or objectively invisible. In the objective condition, visibility was controlled experimentally by fixing the contrast of the masks. In the subjective condition, visibility was controlled by the participant’s response (i.e., visibility became the dependent variable), and mask contrast was continuously adjusted to yield similar proportions of trials rated as subjectively visible and invisible. This approach yielded a similar number of trials for objectively and subjectively invisible stimuli, resulting in similar statistical power to detect neural effects in objective and subjective conditions. Using multivariate pattern analysis, we tracked neural representations of faces and houses along the visual processing hierarchy, from early visual cortex to object-selective lateral occipital complex (LOC [44,45]), as well as in category-selective regions in the lateral occipital (occipital face area [OFA] [46] and occipital place area [OPA] [47]) and ventrotemporal cortices (fusiform face area [FFA] [48] and parahippocampal place area [PPA] [48]). This analysis allowed us to establish the level of representation of objectively and subjectively invisible stimuli in human visual cortex.

Results

On every trial, participants discriminated between faces and houses (objective measure, face/house) and simultaneously indicated stimulus visibility (subjective measure, visible/invisible). Participants were asked to be as accurate as possible, guessing when necessary. For the subjective measure, instructions emphasized that participants should press “visible” even when they had only a vague idea of the stimulus category and press “invisible” only when they had absolutely no idea of the stimulus category. Faces and houses were presented for 16.7 ms, sandwiched between masks (Fig 1A). Subjective stimulus visibility was based on the participant’s response, while objective stimulus visibility was controlled by the experimenter. In the subjective condition (50% of trials), mask contrast was adjusted through an adaptive 1-up 1-down “staircase” procedure. Following a “visible” response, mask contrast was increased by 4%. Following an “invisible” response, contrast was lowered by 4%. This continuous adjustment was intended to yield a similar number of subjectively visible (subj-vis) and subjectively invisible (subj-inv) trials. This allowed us to test the effect of subjective visibility with minimal differences in mask contrast (subj-vis: mean contrast 7.8%, SD 4.4; subj-inv: mean contrast 11.9%, SD 4.3). In the objective condition (50% of trials), mask contrast was either low (2%) to achieve clear stimulus visibility (objectively visible (obj-vis) condition) or high (100%) to achieve chance-level discrimination (objectively invisible (obj-inv) condition).

Experimental paradigm, results from the behavioral masking efficiency experiment and behavioral results from the fMRI experiment.
Fig 1
(a) Example trial: A face or house stimulus was forward and backward masked by a mask with variable contrast, depending on the visibility condition. On every trial, participants judged both the category and visibility of the masked stimulus, yielding measures of objective category discriminability and of subjective stimulus visibility. (b) Results from the behavioral masking efficiency experiment. Mean face-house discriminability is plotted for 9 different masking contrast levels (on a log10 scale). Only with 100% masking contrast did discriminability not significantly exceed chance performance. (c) Mean visibility ratings from the behavioral masking efficiency experiment. (d) Behavioral results from the fMRI experiment. Mean face-house discriminability is shown for the four visibility conditions. Mask contrast was set to 2% in obj-vis and to 100% in obj-inv; in the subjective conditions, mask contrast was continuously adjusted by an adaptive staircase procedure, which resulted in a mean contrast (across the group) of 8% in subj-vis, and of 12% in subj-inv. Discrimination performance was significantly above chance for subjectively invisible stimuli (subj-inv) but not for objectively invisible stimuli (obj-inv). All error bars represent 95% confidence intervals; in (d), for the obj-inv condition, the error bar was smaller than the symbol. (e) Individual participant’s face-house discrimination performance in the obj-inv condition (in d’). Every gray circle represents a participant, and the red circle shows the group mean with its 95% confidence interval. Data underlying this figure are available on OSF (https://osf.io/qus5v/). fMRI, functional magnetic resonance imaging; obj-inv, objectively invisible; obj-vis, objectively visible; subj-inv, subjectively invisible; subj-vis, subjectively visible.Experimental paradigm, results from the behavioral masking efficiency experiment and behavioral results from the fMRI experiment.

Masking efficiency experiment

We first ran a separate behavioral experiment to measure the influence of mask contrast on perceptual discriminability and to determine optimal mask contrast for objective invisibility (N = 17, Fig 1B). Face/house discriminability and subjective visibility (Fig 1C) increased similarly with decreasing masking strength (F(8, 128) = 247.15, p < 0.001, ηp2 = 0.94, BF10 = 2.07 × 1074 and F(8, 128) = 215.69, p < 0.001, ηp2 = 0.93, BF10 = 8.99 × 1069, respectively; also see S1 Fig). A set of one-tailed t tests showed significant above-chance performance for all mask contrasts (all p < 0.005, all dz > 0.75, all BF+0 > 14), except for 100% mask contrast (M = 0.03, SD = 0.41, t(16) = 0.26, p = 0.40, dz = 0.06, BF0+ = 3.26). Bayes factor (BF0+ ) (note that BF was calculated in JASP [93] with default prior scales [Cauchy distribution, scale 0.707]; see S1 Text and S2 Fig for additional analyses of discrimination performance in the obj-inv condition) indicated that the null hypothesis of chance performance was about three times more likely than the alternative hypothesis of above-chance performance, which represents “moderate” evidence for the null hypothesis [49]. Thus, full mask contrast was necessary to achieve objective invisibility and was therefore used in the following fMRI experiment. Note that optimal mask contrast was determined based on the group data from the masking efficiency experiment and set accordingly for the whole group of fMRI participants. An alternative approach that is sometimes thought to be better suited to localize the “sweet spot” for unconscious processing is to calibrate mask contrast individually to allow for maximum signal strength for every participant [50]. However, the masking efficiency experiment revealed low between-subject variability, as well as low reliability for the estimation of individual performance measures for high mask contrasts (see S1 Text and S1 Table). We thus considered group-based calibration of mask contrast better suited to ensure chance performance for all participants in the fMRI experiment, although we cannot exclude the possibility that our group-based calibration approach did not result in optimal stimulus strength for every individual observer. This might have reduced the chances of finding higher-level processing for objective invisibility and might have inflated the difference between subjective and objective invisibility in our study.

Behavior

In the fMRI experiment (N = 43), face/house discriminability (d’) was high in obj-vis (M = 4.50, SD = 0.63) and in subj-vis trials (M = 3.50, SD = 0.93) and remained above chance in subj-inv (M = 1.20, SD = 0.94, t test against chance, t(42) = 8.36, p < 0.001 (one-tailed), dz = 1.26, BF+0 = 8.59 × 109, Fig 1D. Such above-chance performance for subjectively invisible stimuli is often referred to as a blindsight-like phenomenon of unconscious stimulus processing. Importantly, in obj-inv trials, discrimination performance (M = 0.02, SD = 0.25) did not differ significantly from chance (Fig 1E), with moderate evidence for the null hypothesis of chance-level discrimination (t(42) = 0.45, p = 0.33 (one-tailed), dz = 0.07, BF0+ = 4.13; also see S2 Fig). Regarding subjective visibility, participants reported that the stimulus was visible in 94.6% (SD 5.5) of the obj-vis trials and invisible in 96.3% (SD 5.1) of the obj-inv trials. In the subj-vis and subj-inv trials, mean subjective visibility was by definition 100% and 0%, respectively, because these trial categories were conditioned on the subjective visibility response.

Category-specific information in visual cortex

We tested in which subregions of visual cortex activation patterns reliably distinguished between faces and houses. To obtain a measure of category information in the neural responses, activity patterns were correlated with “benchmark” patterns for unmasked faces and houses obtained in an independent localizer scan where participants did a simple one-back task on stimuli presented in a block design. This approach is statistically powerful and rules out the possibility that multivoxel pattern correlations reflect cognitive processes related specifically to the task or decision and motor processes (as these differed between the localizer and the main experiment) [51,52]. Category information was quantified as the difference between within- and between-category multivoxel pattern correlations (Δr) between the independent localizer and every visibility condition in the main experiment for 4 separate regions of interest (ROIs, see Fig 2). Note that differences in multivoxel pattern correlations can result from changes in the fine-grained activation patterns within the face- and house-selective regions or from univariate activation differences between these regions (for complimentary univariate analyses, see S1 Text and S3 Fig).

Category information across regions in visual cortex.
Fig 2
Multivoxel pattern correlations between the 4 visibility conditions in the main experiment and the independent localizer. Bars show the mean difference between within-category and between-category multivoxel pattern correlations (Δr). Error bars represent 95% confidence intervals. (a) Results for early visual cortex (V1) vs. object-selective visual cortex (LOC). (b) Results for posterior category-selective areas (OFA/OPA) vs. anterior category-selective areas (FFA/PPA). Data underlying this figure are available on OSF (https://osf.io/qus5v/). FFA, fusiform face area; LOC, lateral occipital complex; OFA, occipital face area; OPA, occipital place area; PPA, parahippocampal place area; V1, primary visual cortex.Category information across regions in visual cortex.

First, we examined activity patterns in primary visual cortex (V1) and object-selective visual cortex (LOC [44]). There were marked differences between the regions, with overall greater category information in LOC than in V1 (F(1, 42) = 260.39, p < 0.001, ηp2 = 0.86, BF10 = 4.53 × 1031). Category information also differed between visibility conditions (F(3, 126) = 79.03, p < 0.001, ηp2 = 0.65, BF10 = 1.58 × 1028), and these differences were more pronounced in LOC than in V1 (interaction, F(3, 126) = 99.86, p < 0.001, ηp2 = 0.70, BF10 = 1.48 × 1030). To test whether activity patterns discriminated between faces and houses with above-chance accuracy, separate one-tailed t tests were carried out for every region and every visibility condition. V1 carried significant category information in obj-vis (t(42) = 3.44, p < 0.001, dz = 0.52, BF+0 = 46.73), subj-vis (t(42) = 3.97, p < 0.001, dz = 0.60, BF+0 = 188.02), and obj-inv (t(42) = 2.55, p = 0.007, dz = 0.39, BF+0 = 5.74), but not in subj-inv (t(42) = 1.12, p = 0.135, dz = 0.17, BF0+ = 1.97). Activity patterns in LOC discriminated between faces and houses with above-chance accuracy in all visibility conditions (obj-vis (t(42) = 16.69, p < 0.001, dz = 2.55, BF+0 = 2.43 × 1017), subj-vis (t(42) = 12.88, p < 0.001, dz = 1.96, BF+0 = 3.52 × 1013), subj-inv (t(42) = 5.21, p < 0.001, dz = 0.80, BF+0 = 7.22 × 103), and also in obj-inv (t(42) = 3.28, p = 0.001, dz = 0.50, BF+0 = 31.19). Thus, LOC contained category information for both subjectively and objectively invisible stimuli.

To gain insight into the level of representation in higher-level visual cortex, we examined activity patterns in posterior versus anterior category-selective regions, which are partially overlapping with the spatially extended LOC region (Fig 2 ). The 2 more posterior areas represent more basic and “local” visual aspects of faces and houses (OFA [46] and OPA [47,53]), whereas the two more anterior areas also contain “global” and abstract categorical representations (FFA and PPA [48]). Overall, anterior regions carried more category information than posterior regions (F(1, 42) = 47.09, p < 0.001, ηp2 = 0.53, BF10 = 4.53 × 1031). There were also significant differences between visibility conditions (F(3, 126) = 63.76, p < 0.001, ηp2 = 0.60, BF10 = 1.58 × 1028) and a significant interaction (F(3, 126) = 26.94, p < 0.001, ηp2 = 0.39, BF10 = 1.48 × 1030). Activity patterns in the posterior areas (OFA/OPA) discriminated between faces and houses in all conditions (obj-vis [t(42) = 10.55, p < 0.001, dz = 1.61, BF+0 = 6.99 × 1010], subj-vis [t(42) = 6.99, p < 0.001, dz = 1.07, BF+0 = 1.72 × 106], subj-inv [t(42) = 3.90, p < 0.001, dz = 0.59, BF+0 = 155.49], and obj-inv [t(42) = 2.72, p = 0.005, dz = 0.42, BF+0 = 8.27]). Also in the anterior areas (FFA/PPA) category information was significant in all visibility conditions. Strong evidence for above-chance discrimination of faces and houses was obtained only for obj-vis (t(42) = 11.43, p < 0.001, dz = 1.74, BF+0 = 7.89 × 1011), subj-vis (t(42) = 9.61, p < 0.001, dz = 1.47, BF+0 = 4.86 × 109), and subj-inv (t(42) = 2.72, p = 0.005, dz = 0.90, BF+0 = 5.88 × 104). By contrast, for obj-inv, there was no solid evidence for above-chance discrimination in FFA/PPA, with only weak and inconsistent statistical evidence (t(42) = 1.97, p = 0.028, dz = 0.30, BF+0 = 1.86).

Differences between visibility conditions

To directly test the effect of different definitions of (in)visibility, we compared activity patterns using paired t tests (two-tailed). Both in V1 and LOC, subjective visibility was associated with greater category information than subjective invisibility (V1: t(42) = 2.85, p = 0.007, dz = 0.43, BF10 = 5.57; LOC: t(42) = 9.68, p < 0.001, dz = 1.48, BF10 = 3.00 × 109), but this effect was larger in LOC (interaction, F(1, 42) = 62.93, p < 0.001, ηp2 = 0.60, BF10 = 7.99 × 105), indicating a greater effect of subjective awareness in object-selective visual cortex than in early visual cortex. Also the way of establishing invisibility influenced the two regions differently (interaction, F(1, 42) = 9.02, p = 0.004, ηp2 = 0.18, BF10 = 11.14). While there was no significant difference in category information between subjectively and objectively invisible stimuli in V1 (t(42) = −0.73, p = 0.47, dz = 0.11, BF01 = 4.72), in LOC category information was significantly greater for subjectively than for objectively invisible stimuli (t(42) = 2.67, p = 0.011, dz = 0.41, BF10 = 3.72).

Similar differences were obtained for posterior versus anterior category-selective regions. Although both posterior and anterior regions carried more category information for subjectively visible than for subjectively invisible stimuli (OFA/OPA: t(42) = 4.24, p < 0.001, dz = 0.64, BF10 = 195.43; FFA/PPA: t(42) = 6.42, p < 0.001, dz = 0.98, BF10 = 1.49 × 105), this effect was larger in FFA/PPA (interaction, F(1, 42) = 11.41, p = 0.002, ηp2 = 0.21, BF10 = 5.66), indicating a greater effect of subjective awareness in anterior than in posterior category-selective regions. Critically, also the method of establishing invisibility had a different effect on the two ROIs (interaction, F(1, 42) = 20.43, p < 0.001, ηp2 = 0.33, BF10 = 28.18). In OFA/OPA, there was no significant difference between subjectively and objectively invisible stimuli (t(42) = 0.79, p = 0.44, dz = 0.12, BF01 = 4.50), while FFA/PPA carried more category information for subjectively than for objectively invisible stimuli (t(42) = 4.23, p < 0.001, dz = 0.64, BF10 = 195.71).

Posterior–anterior category information gradient

Another way to follow up on the significant visibility × ROI interactions is to test differences in category information between ROIs, separately for every visibility condition. For obj-vis, subj-vis, and subj-inv, this analysis revealed a posterior–anterior gradient, with increasing category information from V1 to LOC (all t(42) > 4.50, p < 0.001, dz > 0.70, BF10 > 596.40, Fig 2A) and from OFA/OPA to FFA/PPA (all t(42) > 4.45, p < 0.001, dz > 0.67, BF10 > 373.24, Fig 2B). For obj-inv, this gradient was not robust. Category information did not differ between V1 and LOC (t(42) = 1.70, p = 0.097, dz = 0.26, BF01 = 1.63) or between OFA/OPA and FFA/PPA (t(42) = 0.20, p = 0.842, dz = 0.03, BF01 = 5.95), indicating that processing of objectively invisible stimuli is limited to visual shape features.

Whole-brain searchlights

The ROI analyses showed that both subjective visibility (subj-vis vs. subj-inv) and the method for establishing invisibility (subj-inv vs. obj-inv) have the strongest effects in higher-level visual cortex. To further substantiate these findings and to test their spatial specificity, we conducted “searchlight” analyses [54]. These analyses revealed areas carrying significant category information across the whole brain (corrected for multiple comparisons via false discovery estimation [55], p < 0.05) in the obj-vis, subj-vis, and subj-inv condition, but not in the obj-inv condition (Fig 3A). In the three former visibility conditions, clusters with significant category information were located in bilateral fusiform gyrus, lingual gyrus, parahippocampal gyrus, and inferior occipital gyrus—areas overlapping with the high-level visual cortical areas defined in our ROI analyses. Compared to subjectively invisible stimuli, subjectively visible stimuli were associated with greater category information in the very same regions, and in bilateral fusiform gyrus in particular, thus confirming the results from the ROI analyses (Fig 3B). Additional clusters for subjectively visible stimuli were located in bilateral inferior temporal gyrus, mid-occipital gyrus, superior occipital gyrus, left superior parietal lobule, precuneus, and left amygdala. Compared to this effect of subjective visibility, the contrast of objectively visible with objectively invisible stimuli revealed greater and more widespread differences in category information in overlapping occipitotemporal regions, as well as additional effects in inferior frontal regions. When directly contrasting subjectively invisible with objectively invisible stimuli, there was greater category information for subjectively invisible stimuli in right fusiform gyrus and in left anterior fusiform gyrus extending into left inferior temporal gyrus (Fig 3B).

(a) Searchlight results showing clusters with significant category information across the whole brain, separately for the different visibility conditions. Only voxels surviving multiple comparison correction via false discovery estimation [55] (p < 0.05 FDR corrected) are shown. For the obj-inv condition, no clusters survived this statistical threshold. (b) Searchlight results showing the effect of subjective visibility by comparing the obj-vis to the obj-inv condition, the subj-vis to the subj-inv condition, and the effect of method for establishing invisibility by comparing the subj-inv to the obj-inv condition. Data underlying this figure are available on OSF (https://osf.io/qus5v/). FDR, false discovery rate; obj-inv, objectively invisible; obj-vis, objectively visible; subj-inv, subjectively invisible; subj-vis, subjectively visible.
Fig 3
(a) Searchlight results showing clusters with significant category information across the whole brain, separately for the different visibility conditions. Only voxels surviving multiple comparison correction via false discovery estimation [55] (p < 0.05 FDR corrected) are shown. For the obj-inv condition, no clusters survived this statistical threshold. (b) Searchlight results showing the effect of subjective visibility by comparing the obj-vis to the obj-inv condition, the subj-vis to the subj-inv condition, and the effect of method for establishing invisibility by comparing the subj-inv to the obj-inv condition. Data underlying this figure are available on OSF (https://osf.io/qus5v/). FDR, false discovery rate; obj-inv, objectively invisible; obj-vis, objectively visible; subj-inv, subjectively invisible; subj-vis, subjectively visible.

Correlates of subjective awareness

As mask contrast covaried with subjective visibility, stronger category representations in the subj-vis than in the subj-inv condition could reflect differences in both stimulus strength and subjective awareness. To control for the effect of stimulus strength, we modeled brain responses in the subjective condition (separately for faces and houses) with a parametric regressor reflecting trial-by-trial mask contrast and an additional regressor for visibility (subj-vis, subj-inv). The visibility regressor was orthogonalized with respect to the mask-contrast regressor, assigning the mask-contrast regressor all shared variance, such that only variance not explained by the mask-contrast regressor was assigned to the visibility regressor [56]. Note that if fluctuations in subjective awareness and in mask contrast reflected partially shared processes, such as fluctuations in arousal or attention, these would be captured by the mask-contrast regressor, rendering this analysis a conservative approach for determining the neural correlates of subjective awareness.

Multivoxel pattern correlations between the localizer data and the face-/house patterns adjusted for mask contrast (Fig 4A) revealed strong effects of subjective visibility (one-sample t test, two-tailed) in the more anterior areas LOC (t(42) = 3.57, p < 0.001, dz = 0.54, BF10 = 32.03) and FFA/PPA (t(42) = 3.08, p = 0.004, dz = 0.47, BF10 = 9.45), but only inconsistent effects in the more posterior areas V1 (t(42) = 1.96, p = 0.056, dz = 0.30, BF10 = 0.94) and OFA/OPA (t(42) = 2.21, p = 0.033, dz = 0.34, BF10 = 1.46). Note, however, that there were no significant differences between posterior and anterior ROIs (both t(42) < 1.37, p > 0.17, dz < 0.21, BF01 > 2.57), limiting claims about the spatial specificity of the effect. To determine whether subjective awareness enhanced category information specifically in occipitotemporal regions, we conducted a whole-brain searchlight analysis using the model controlling for mask contrast. Even at a more liberal statistical threshold (p < 0.001, uncorrected) this analysis revealed larger clusters with greater category information in subj-vis than in subj-inv only in bilateral fusiform gyrus (Fig 4B) and in bilateral mid-occipital gyrus (additional small clusters were located in superior parietal lobule). This further highlights the spatial specificity of the effect of subjective awareness.

Effect of subjective awareness, adjusted for differences in mask contrast between the subj-vis and subj-inv condition.
Fig 4
(a) Difference in multivoxel pattern correlations (Δr) between the subj-vis and subj-inv condition in the four ROIs. *p < 0.05 **p < 0.01 ***p < 0.001. (b) Searchlight results showing clusters with significantly greater category information in the subj-vis than in the subj-inv condition across the whole brain, controlling for mask contrast. Note that this map is thresholded at p < 0.001, uncorrected. Data underlying this figure are available on OSF (https://osf.io/qus5v/). FFA, fusiform face area; LOC, lateral occipital complex; OFA, occipital face area; OPA, occipital place area; PPA, parahippocampal place area; ROI, region of interest; subj-inv, subjectively invisible; subj-vis, subjectively visible; V1, primary visual cortex.Effect of subjective awareness, adjusted for differences in mask contrast between the subj-vis and subj-inv condition.

Correlates of blindsight-like discrimination performance

To test for brain areas involved in blindsight-like unconscious perception [25], we harnessed the fact that there was considerable interindividual variability in behavioral discriminability of subjectively invisible stimuli (Fig 5A ). The phenomenon of blindsight refers to above-chance behavioral performance in the absence of subjective awareness of the stimulus on which these behavioral effects are based [57]. For the subj-inv condition, correlations between perceptual discriminability (d’, behavioral performance) and category information in ROIs (Δr) were significant in LOC (p < 0.001), OFA/OPA (p = 0.028) and FFA/PPA (p = 0.001), but not in V1 (p = 0.88). One concern with these analyses is that mask contrast in the subj-inv condition was adjusted dynamically for each observer and thus differed between participants. Indeed, mask contrast was (negatively) correlated with behavioral performance (r(41) = −0.40, p = 0.008, BF10 = 5.74). We therefore computed partial correlations between perceptual discriminability and the ROI data, controlling for the effect of mask contrast. As can be seen in Fig 5B, also with this analysis there were significant correlations between discrimination performance and category information in LOC (r(41) = 0.38, p = 0.013, BF10 = 4.68) and in FFA/PPA, (r(41) = 0.38, p = 0.012, BF10 = 4.89), but the correlation with OFA/OPA was not significant anymore (r(41) = 0.29, p = 0.064, BF10 = 1.43). To examine the spatial specificity of these correlations, we calculated a correlation between perceptual discriminability and the subj-inv whole-brain searchlight. Even at a liberal statistical threshold (p < 0.001, uncorrected, Fig 5C), this analysis only revealed a cluster in left fusiform gyrus, further indicating the spatial specificity of the results.

Blindsight-like discrimination of subjectively invisible stimuli.
Fig 5
(a) Perceptual discrimination of subjectively invisible stimuli (d’) was highly variable between individuals. Every circle represents a participant; the horizontal bar the mean, and the error bar the 95% confidence interval. (b) Multiple regression analyses of category information (Δr) in the four ROIs for the subj-inv condition, showing standardized beta weights for the predictors perceptual discrimination of subjectively invisible stimuli (d’) and mask contrast in the subj-inv condition. Error bars show 95% confidence intervals. *p < 0.05 **p < 0.01. (c) Partial correlation between perceptual discrimination and the subj-inv whole-brain searchlight, adding the effect of mask contrast to the null model. Note that this map is thresholded at p < 0.001, uncorrected. Data underlying this figure are available on OSF (https://osf.io/qus5v/). FFA, fusiform face area; LOC, lateral occipital complex; OFA, occipital face area; OPA, occipital place area; area; PPA, parahippocampal place area; ROI, region of interest; subj-inv, subjectively invisible; V1, primary visual cortex.Blindsight-like discrimination of subjectively invisible stimuli.

Discussion

In the absence of a gold standard for measuring consciousness, scientists are firmly divided into those who use subjective measures based on participant’s reported experience and those who use objective measures based on participant’s perceptual discrimination performance [11,16]. Using fMRI, we found that both objective and subjective visibility were associated with a clear posterior-to-anterior gradient: While there was little effect of visibility on category information in early visual cortex, higher-level visual cortex was strongly modulated by visibility. In frontal cortex, we did not obtain evidence for such effects of visibility on category representations. When investigating neural processing of invisible stimuli, subjective measures allow for greater stimulus strength (e.g., lower mask contrast) than objective measures, and this may result in greater estimates of unconscious processing. We found significant category information even for objectively invisible stimuli in both early, lower-level visual areas and in higher-level visual cortex, although representations in anterior, category-selective ventrotemporal areas (FFA/PPA) were less robust. Subjectively invisible stimuli were more strongly represented in object-selective visual cortex than objectively invisible stimuli. Furthermore, for subjectively invisible stimuli, similar to visible stimuli, there was a clear posterior-to-anterior gradient in visual cortex, with stronger category information in ventrotemporal cortex (FFA/PPA) than in occipital cortex (V1, OFA/OPA). For objectively invisible stimuli, category information remained virtually unchanged from early visual cortex to object- and category-selective areas. We interpret this as an indication that representations of objectively invisible stimuli are limited to visual (shape-related) object properties processed in early, lower-level visual areas, while subjectively invisible stimuli are processed up to more categorical levels of representation in higher-level category-selective areas. Although differences in multivoxel pattern correlations could reflect differences in response amplitude or changes in activation patterns, we obtained almost identical results in a univariate ROI analysis (see S1 Text and S3 Fig ), and thus consider differences in response amplitude between face- and house-selective regions the more likely account for differences between visibility conditions. These differences in neural processing of subjectively versus objectively invisible stimuli can account for some of the current controversies on the scope and extent of unconscious processing [9,14].

Unconscious processing of objectively invisible stimuli has been notoriously difficult to demonstrate. Rendering stimuli perceptually indiscriminable through visual masking requires very low stimulus strength (e.g., high mask contrast), which strongly reduces neural responses in visual cortex, especially at higher levels in the cortical hierarchy [5]. Furthermore, above-chance performance in objective measures may be driven by unconscious processing [43]. As a result, the use of objective measures may underestimate unconscious processing. Indeed, several previous fMRI studies failed to obtain visual cortex responses to invisible faces [38,58], shapes [59], and objects [25,60]. Here, although maximum mask contrast was required for objective invisibility, perceptually indiscriminable faces and houses still evoked distinct activity patterns in visual cortex. Category information could be decoded from early visual cortex, from object-selective LOC, from category-selective areas in lateral occipital cortex (OFA/OPA), and, less robustly, from ventrotemporal areas (FFA/PPA), while participants were unable to discriminate these categories, thereby satisfying the most rigorous criteria for establishing absence of awareness (see also S1 Text and S2 Fig for more details on this issue). Previous fMRI studies that found visual cortex responses to masked stimuli did not convincingly rule out conscious processing because these studies had very small sample sizes [3338] and measured awareness with fewer trials [35,37,3942] (or even fewer participants [38,40]), often in a separate block outside the scanner [3541]. With low power, a nonsignificant effect in the awareness measure (p > 0.05) is not surprising and hard to interpret. Even if both the awareness measure and the fMRI data reflected the same underlying (conscious) process, failure to reject the null hypothesis is expected to occur more frequently for the measure with lower power (the awareness measure). To address these concerns, we tested a much larger sample than previous fMRI studies on unconscious processing, collected objective awareness measures on all 200 trials of the objectively invisible condition during scanning, and calculated BF to quantify the evidence favoring the null hypothesis of zero face/house discriminability versus the alternative hypothesis of above-chance discrimination performance [61]. Finally, no participants or trials needed to be excluded based on high awareness scores. Such data exclusion is a common procedure that risks severely inflating estimates of unconscious processing due to regression to the mean [62]. On a cautionary note, however, although we adopted the highest standards for demonstrating unconscious processing with fMRI to date, our claim of complete absence of awareness depends on specific statistical assumptions such as the prior in Bayesian analyses (also see S1 Text and S2 Fig). Notwithstanding these limitations, our findings provide more conclusive fMRI evidence for unconscious processing of objectively invisible stimuli in human visual cortex.

While representations of objectively invisible stimuli were of similar strength along the posterior–anterior axis from early visual cortex to object- and category-selective areas, subjectively invisible stimuli—similar to visible stimuli—were most strongly represented in category-selective areas in ventrotemporal cortex (FFA/PPA). These differences in neural representation may provide an account for discrepant findings on the influence of unconscious stimuli on behavior, for example in visual priming experiments [13,63,64]. For objectively invisible stimuli, our findings predict priming effects based on visual features such as shape, but not based on semantic meaning such as category membership (e.g., gender, emotion, or animacy). Indeed, while response priming from objectively invisible shapes (e.g., left- versus right-pointing arrow [65] or square versus diamond [66]) is well established, semantic priming from objectively invisible pictures, where visual effects are ruled out (e.g., a picture of an animal priming a word referring to an animal), is heavily debated [13,63,64]. When objective discrimination performance is well above chance, as for subjectively invisible stimuli in the present study, masked primes elicit robust semantic processing [67]. Recent studies indicate that subjectively invisible stimuli reach even higher levels of processing, including crossmodal semantic integration [68] and working memory [69]. The present findings will resonate with the idea that subjectively invisible stimuli are processed in a way that is qualitatively similar to (clearly) visible stimuli [14].

Activity patterns in LOC and FFA/PPA also predicted perceptual discrimination performance for subjectively invisible stimuli, representing a neural correlate of blindsight-like unconscious perception in healthy human observers [25]. For proponents of subjective measures to study unconscious processes, this could be regarded as strong evidence that unconscious perception shares neural mechanisms with conscious perception, providing support for the idea that consciousness has little functional role in human perception and cognition [14]. However, for proponents of objective awareness measures, above-chance performance for subjectively invisible stimuli simply means that these stimuli were in fact not invisible. According to this view, differences in perceptual performance on subjectively invisible trials could reflect differences in response criteria (willingness to say “visible” to barely visible stimuli [26,2932]). In our staircasing procedure, mask contrast was continuously adjusted based on subjective visibility, such that a more liberal criterion (greater willingness to say “visible”) would result in higher average mask contrast and this could account for lower performance. Patterns in LOC and FFA/PPA were better predicted by perceptual performance than by mask contrast, and perceptual performance explained a significant portion of the neural variance taking into account the variance explained by mask strength. However, without a measure of sensitivity and criterion for the visibility task (e.g., by including stimulus-absent trials), our results cannot rule out criterion effects.

Because it is the very nature of subjective measures that visibility judgments are subject to the observer’s own interpretation and response criterion, it is possible that results would have differed for the subjectively invisible condition had we adopted another visibility scale. Although binary visibility scales such as the one used here have been widely adopted [26], for example in studies using masking [70,71], attentional blink [72], dichoptic fusion [73], and interocular suppression [74,75]; recently, the four-point perceptual awareness scale (PAS) [17] has gained in popularity. Depending on specific instructions and observer’s interpretations, the present “invisible” rating may have corresponded to the lowest level or to the two lowest PAS levels. While some PAS studies defined invisibility as the two lowest PAS levels [68,76], using only the lowest level is more common in the literature [20,21,23]. Compared to this approach, our subjective measure may have been more liberal in defining invisibility, resulting in greater estimates of blindsight-like above-chance performance for subjectively invisible stimuli. This highlights a general problem with subjective measures. Subjective states are “private”, i.e., they cannot be externally falsified, so that differences between scales, observers, or studies cannot be unequivocally linked to differences in perceptual states. One solution to this problem is to adopt objective measures, which may come at the cost of failing to capture the subjective quality of conscious experience and which may be influenced by unconscious processes. For these reasons, it is difficult to determine whether above-chance performance for subjectively invisible stimuli reflects unconscious processing or a conservative response criterion in the subjective measure. One promising avenue for future studies is to adopt recently developed criterion-free measures of subjective awareness such as meta-d’ based on confidence ratings [77].

Nevertheless, by comparing stimuli of similar stimulus strength reported as “visible” versus “invisible,” subjective measures have often been adopted to reveal the neural correlates of subjective awareness. Although the contrast of subjectively visible and invisible trials may be confounded with factors unrelated to awareness (e.g., alertness, attention), many previous fMRI studies adopted this approach, often using binocular rivalry or other bistable stimuli, and typically found stronger visual cortex responses for stimuli reported as visible [12]. Our finding of better pattern discrimination of subjectively visible stimuli in object- and category-selective visual cortex is consistent with the idea that subjective awareness is related to enhanced activity in the very same brain areas that are specialized for processing those stimuli [78,79]. Whether activity in this posterior occipito–temporo–parietal “hot zone” [80] is sufficient for awareness, or whether awareness requires additional activity in prefrontal cortex [7,81], is subject of ongoing debate. Recent studies using so-called “no-report” paradigms indicate that activity in frontal cortex may reflect post-perceptual processes related to reporting awareness rather than to content-specific perceptual awareness per se [82]. Although establishing absence of awareness online during imaging required trial-by-trial visibility and discrimination responses, our approach of correlating activity with “benchmark” patterns from an independent localizer scan where participants had a different task reduced the influence of post-perceptual effects (also see S1 Text ) [51,52]. Ruling out task-specific representations in this way, there was no evidence for prefrontal cortex involvement (see S5 and S6 Figs and S1 Text for evidence that activity in a frontoparietal network distinguishes visible versus invisible conditions, irrespective of stimulus category).

The marked differences in neural representation of subjectively versus objectively invisible stimuli revealed here imply that models where “unconscious processing” is seen as a unitary mechanism or concept, without further specifying how absence of awareness was established, will not provide a good fit to the full range of empirical data. Rendering stimuli subjectively or objectively invisible results in different estimates of unconscious information processing in the human brain. We provide the first conclusive evidence that even fully indiscriminable, objectively invisible stimuli can be decoded from patterns of fMRI activity in human visual cortex, albeit at a drastically lower strength than visible stimuli. Previous studies that made such claims [3338] did not convincingly establish absence of awareness during scanning and/or had very low statistical power (small sample sizes), resulting in high probability of false positives in the fMRI data [83] and of false negatives in the awareness measure [84,85]. However, only subjectively invisible, but not objectively invisible, stimuli were increasingly processed along a posterior–anterior gradient, with greater category information in category-selective ventrotemporal cortex than in occipital cortex.

Methods

Participants

The experiments were approved by the University of Amsterdam Ethics Committee (approval numbers 2019-BC-10091 and 2019-BC-10347), and all procedures were conducted according to the principles expressed in the Declaration of Helsinki. Volunteers were recruited from the University of Amsterdam participant pool. Participants were mostly students who received either course credit or a monetary compensation for their participation. All participants reported normal or corrected-to-normal vision, were naïve to the research question, and provided written informed consent. In the fMRI experiment, we scanned 54 participants. This sample size was constrained by the resources provided by an fMRI grant awarded to the first author. We tested as many participants as possible within the allotted scanning time. Eleven participants were excluded from all data analyses: three were excluded because they did not finish the experiment, two because their anatomical scans were corrupted, and six because of failure to follow the instructions inside the scanner (very few button presses, holding down the buttons continuously, or reporting stimulus visibility in almost all trials of the obj-inv condition). The final sample consisted of 43 participants (24 female, mean age 22.7 years, SD 3.9, range 18 to 37 years). In the behavioral masking efficiency experiment, there were 18 participants, one of which was excluded due to a coding error (unbalanced experimental conditions), resulting in a final sample of 17 participants (12 female, mean age 21.8 years, SD 4.7, range 18 to 31 years).

Stimuli

In the scanner, stimuli were presented on a 32-inch LCD screen for MRI (1920 × 1080 pixel resolution, 120-Hz refresh rate) seen from a viewing distance of approximately 150 cm through a mirror mounted on the head coil. The experiment was programmed in MATLAB using the Psychtoolbox [86] functions. Stimulus presentation was synchronized with the 8.3-ms vertical refresh cycle of the screen. Stimuli were ten face photographs of neutral expression from the FACES database [87] and ten house photographs (taken from the front) selected from the internet. This selection was based on informal pilot testing with a larger set of stimulus exemplars. Pilot results indicated that the selected exemplars were best matched on subjective visibility. Photographs were cropped to an oval of 240 × 336 pixels containing only the inner features of faces and houses, converted to grayscale, and the oval was assigned identical mean luminance and contrast (in RGB values M = 127.5, SD = 17.5). The remainder of the stimulus rectangle as well as the screen background were mid-gray (RGB value 127.5). Masks filled the whole stimulus rectangle and consisted of a randomly generated arrangement of overlapping rectangles and—in a lesser number—circles in various sizes and levels of gray. We created a set of 100 masks, from which one forward mask and one backward mask were randomly selected on every trial. The contrast of these masks differed between visibility conditions and varied between 2% and 100%. Throughout the experiment, all stimuli and masks were presented within an 8-pixel wide light gray frame (248 × 344 pixels, RGB value 143).

In the obj-vis condition, mask contrast was set to 2% to achieve clear visibility of the face/house stimuli. In the obj-inv condition, mask contrast was set to 100%, such that participants’ ability to discriminate between faces and houses was expected not to differ significantly different from chance (based on the results from the masking efficiency experiment). In the subjective condition, mask contrast was adjusted through an adaptive 1-up 1-down staircase procedure: On the first trial of each run, mask contrast started at 18%; following a “visible” response, mask contrast on the next trial was increased by 4%; following an “invisible” response, mask contrast was lowered by 4% (minimum contrast 2%, maximum contrast 100%). This adjustment was intended to yield a roughly similar number of subj-vis trials and subj-inv trials.

Procedure

On every trial, a forward- and backward-masked face or house stimulus was presented, and participants indicated stimulus category and visibility. Every trial began with 292 ms of fixation on, followed by 392 ms fixation off, 92 ms forward mask, 16.7 ms face/house stimulus (two screen refresh cycles), and 192 ms backward mask. Next, a response screen prompted participants to use one of four buttons to indicate stimulus category and visibility, using their left hand for the two left buttons and their right hand for the two right buttons (using a button box inside the scanner and a standard keyboard outside the scanner). The left-most button represented “face, visible,” and the other button on the left represented “face, invisible.” The right-most button represented “house, visible,” and the other button on the right “house, invisible.” Pilot experiments had shown that this compound response and response mapping was intuitive, and all participants received extensive training in using the buttons. There was a response window of 1.8 s in which participants could enter their response. The trial ended with a fixation period of variable (jittered) duration (selected from a uniform distribution between 100 and 900 ms with as many values as trials per fMRI run or as trials per behavioral experiment). The interstimulus interval ranged between 2.9 and 3.7 s (while brain activity was measured with a fast echo-planar imaging [EPI] sequence with a repetition time of 1.6 s; see below).

Instructions

Before beginning the experiment, participants received detailed written and verbal instructions. They were informed that they would be presented with pictures of faces and houses and that masking would be used to degrade visibility of these stimuli, such that some of the stimuli would be visible and some would be invisible. They were instructed to indicate both stimulus category and visibility using the compound response. Participants were informed that when they had absolutely no idea of what category the stimulus represented (i.e., if they did not see anything that indicated that the picture was a face or a house), they should indicate “invisible” and take a guess. If they had some (vague) idea of what the stimulus category could be, they should indicate “visible.” Participants were informed about the 1.8-s response window, but instructions emphasized that (after practice) this would be more than enough time to provide an accurate response, such that there was no speed pressure and that responses should be as accurate as possible.

Before entering the scanner, participants completed one practice run at a computer outside the scanner. At the beginning of the fMRI session, an anatomical scan was acquired, and participants completed another practice run, which was followed by five runs of the main experiment. fMRI runs started and ended with 6.4 s of fixation. In the following 160 trials of a run, there were 40 trials of the obj-vis condition, 40 trials of the obj-inv condition, and 80 trials of the subjective condition (here, the number of trials in the subj-vis and subj-inv condition depended on the participant’s response). Within each condition, each combination of two stimulus categories (faces and houses) and ten stimulus exemplars occurred equally often. Trial order was randomized. In total, there were 200 trials of the obj-vis condition, 200 trials of the obj-inv condition, and 400 trials of the subjective condition.

Localizer run

At the end of the fMRI session, we acquired a functional localizer scan to localize face- and house-responsive voxels in visual cortex. In the localizer, the same stimulus exemplars as in the main experiment were presented in a standard design for localizing category-selective brain areas: Faces and houses were displayed in separate 16-s blocks, where a series of 16 faces or houses was presented unmasked for 750 ms each, followed by 250-ms fixation. There were 20 face blocks and 20 house blocks in alternating order. After every four blocks, there were 16 s of fixation. Within a block, stimulus exemplars were randomly ordered with the constraint that the same exemplar could only be presented two times per block. Participants were instructed to press a button when there was a repetition of an exemplar (we did not record these button presses).

fMRI acquisition

MRI data were collected using a 3 Tesla Philips Achieva MRI scanner (Philips, Eindhoven, the Netherlands) with a 32-channel head coil. At the beginning of the fMRI session, an anatomical scan was acquired using a T1-weighted gradient-echo sequence (220 slices, 1-mm isotropic voxels). Functional images were acquired using a T2*-weighted multiband EPI sequence (56 slices, flip angle 70°, TR 1600 ms, TE 30 ms, 2-mm isotropic voxels). During the main experiment, 348 volumes were recorded per run (for a total of five runs lasting 9:30 min each), and 248 volumes during the localizer run (lasting 6:30 min).

Behavioral masking efficiency experiment

The masking efficiency experiment was conducted to determine the required masking settings to achieve objective invisibility, i.e., a masking setting at which participants’ ability to discriminate between faces and houses was not significantly above chance. Stimuli were presented on a 24-inch LCD screen (1920 × 1080 pixel resolution, 120-Hz refresh rate) seen from a free viewing distance of approximately 80 cm. The masking efficiency experiment was identical to the fMRI experiment, except that nine fixed masking strengths (log10-scale between 2% and 100% contrast) were used. There were 540 trials, in which each combination of two stimulus categories (faces and houses), ten stimulus exemplars, and nine masking strengths (2, 3.3, 5.3, 8.7, 14.1, 23.1, 37.6, 61.3, 100%) occurred equally often. Trial order was randomized, and there were three obligatory breaks.

Analyses

Behavioral data

Trials with no response (on average less than 3%) were excluded from analyses of behavior and fMRI. We calculated the signal detection theory (SDT) sensitivity index d’ as a measure of objective perceptual discriminability of faces and houses: “Face” responses were coded as hits in face trials and as false alarms in house trials. Hit and false alarm rates of 0 or 1 were converted to 1/(2N) and 1−1/(2N), respectively, with N being the number of trials on which the rates were based [27]. Finally, the z-transformed false alarm rate was then subtracted from the z-transformed hit rate to yield d’.

fMRI data preprocessing

Neuroimaging data were preprocessed using fMRIPrep 1.3.2 [88]. For the structural images, preprocessing steps included intensity non-uniformity correction, skull stripping, surface reconstruction, spatial normalization to the ICBM 2009c Nonlinear Asymmetrical template version 2009c [89] (with nonlinear registration), and brain tissue segmentation. Preprocessing steps for the functional images included susceptibility distortion correction, co-registration to the structural image, estimation of head motion parameters, resampling of the BOLD time series to the template, and high-pass filtering (using a discrete cosine filter with 128 s cut-off).

fMRI data modeling

SPM12 was used to fit a general linear model (GLM) to the data from the localizer run and to the data for each run of the main experiment. For the localizer, the GLM contained two regressors (faces and houses) and six regressors of no interest (head motion parameters). For each run of the main experiment, the GLM contained eight regressors (four visibility conditions; objectively and subjectively visible and invisible, and two stimulus categories; faces and houses), as well as six motion regressors. Regressors were convolved with a standard hemodynamic response function, as included in SPM12. The resulting beta weights for each voxel were used as the data points for the following analyses.

Multivoxel pattern analyses

Multivoxel pattern analysis (MVPA; using the CoSMoMVPA toolbox [90]) was used to identify brain regions carrying category-specific information in their activity patterns, i.e., activity that reliably distinguished between face and house stimuli. Activity patterns were correlated between the block-design localizer and the event-related main experiment. For every participant and for each of the four visibility conditions (obj-vis, subj-vis, subj-inv, and obj-inv), betas from the localizer for one stimulus category (faces and houses) were correlated with betas from the main experiment for the same stimulus category, yielding within-category correlations, and with betas from the main experiment for the other stimulus category, yielding between-category correlations. Correlations were then Fisher z-transformed, the two within-category correlations were averaged, and the two between-category correlations were averaged. Finally, between-category correlations were subtracted from within-category correlations, yielding a correlation difference Δr for each visibility condition [51,52]. Positive correlation differences indicate that activity patterns carry information about stimulus category.

Region of interest analyses

Pattern correlation differences were calculated for four different (bilateral) ROIs. A probabilistic atlas of retinotopic cortex [91] was used to define bilateral early visual cortex (V1). A functional group atlas [92] was used to define bilateral LOC, posterior category-selective areas (face-selective OFA and scene-/house-selective OPA [transverse occipital sulcus]), and anterior category-selective areas (face-selective FFA and scene-/house-selective PPA). For every participant, the localizer data were used to select the 100 most face-responsive and the 100 most house-responsive voxels within each of these ROIs (based on a t test comparing face and house responses; for similar results with other ROI definitions, see S4 Fig and S1 Text).

Searchlight analyses

We additionally ran searchlight analyses [54] to identify activity patterns that distinguished between categories across the whole brain. For these searchlights, we repeatedly calculated pattern correlation differences in the same way as described for the ROI analyses but for a moving sphere with a radius of five voxels (524 voxels in each sphere) which was centered on every voxel in the functional images of every participant (for searchlight analyses based on the data from the main experiment only, see S1 Text and S5 and S6 Figs).

Controlling for mask contrast in the subjective condition

To control for trial-by-trial differences in mask contrast in the subjective condition, another GLM including a parametric regressor for mask contrast and an additional regressor for visibility, separately for faces and houses, was fit to the data from the subjective condition (all other aspects of the GLM were the same as described above). The visibility regressor was orthogonalized with respect to the mask-contrast regressor, such that the visibility regressor was assigned only the variance not explained by the mask-contrast regressor. Both the mask-contrast and the visibility regressor were mean-centered (run-wise, separately for faces and houses).

Statistics

For behavior and ROI data, we report both standard frequentist statistics and BFs calculated in JASP [93] with default prior scales (Cauchy distribution, scale 0.707). When frequentist statistics indicate a significant effect, the corresponding BF is reported as a quantification of the evidence for the alternative hypothesis (BF10); when the effect is not significant, the reported BF quantifies the evidence for the null hypothesis (BF01). To demonstrate absence of awareness in our objective measure, the directional BF0+ quantifies evidence for null sensitivity compared to the alternative of above-chance performance. For multifactorial ANOVAs, we report the inclusion BF quantifying the evidence for all models containing a particular effect compared to all models without that effect. To test for category information in ROIs, the directional BF+0 quantified evidence for above-chance information compared to the alternative of zero information. For the searchlight group maps, results were corrected for multiple comparisons via false discovery rate (FDR) corrections [55] (p < 0.05).

Acknowledgements

We thank T. Beemsterboer, T. Derckx, and S. Rasche for help with data collection and L. Snoek for implementing the standardized “fMRIPrep” pipeline.

Abbreviations

BFBayes factor
EPIecho-planar imaging
FDRfalse discovery rate
FFAfusiform face area
fMRIfunctional magnetic resonance imaging
GLMgeneral linear model
LOClateral occipital complex
MVPAmultivoxel pattern analysis
obj-invobjectively invisible
obj-visobjectively visible
OFAoccipital face area
OPAoccipital place area
PASperceptual awareness scale
PPAparahippocampal place area
ROIregion of interest
SDTsignal detection theory
subj-invsubjectively invisible
subj-vissubjectively visible
V1primary visual cortex

References

DehaeneS, LauH, KouiderS. What is consciousness, and could machines have it? Science. 2017;358:48692. doi: 10.1126/science.aan8871

LeDouxJE, MichelM, LauH. A little history goes a long way toward understanding why we study consciousness the way we do today. Proc Natl Acad Sci U S A. 2020;117:697684. doi: 10.1073/pnas.1921623117

SethA, ConsciousnessK. The last 50 years (and the next). Brain Neurosci Adv. 2018;2:16. doi: 10.1177/2398212818808710

DoerigA, SchurgerA, HerzogMH. Hard criteria for empirical theories of consciousness. Cogn Neurosci. 2021;12:4162.

DehaeneS, ChangeuxJP, NaccacheL, SackurJ, SergentC. Conscious, preconscious, and subliminal processing: A testable taxonomy. Trends Cogn Sci. 2006;10:20411. doi: 10.1016/j.tics.2006.03.007

LammeVAF. Why visual attention and awareness are different. Trends Cogn Sci. 2003;7:128. doi: 10.1016/s1364-6613(02)00013-x

LauH, RosenthalD. Empirical support for higher-order theories of conscious awareness. Trends Cogn Sci. 2011;15:36573. doi: 10.1016/j.tics.2011.05.009

TononiG, BolyM, MassiminiM, KochC. Integrated information theory: From consciousness to its physical substrate. Nat Rev Neurosci. 2016;17:45061. doi: 10.1038/nrn.2016.44

HesselmannG, MoorsP. Definitely maybe: Can unconscious processes perform the same functions as conscious processes? Front Psychol. 2015;6:584. doi: 10.3389/fpsyg.2015.00584

10 

MoorsP, HesselmannG, WagemansJ, van EeR. Continuous flash suppression: Stimulus fractionation rather than integration. Trends Cogn Sci. 2017;21:71921. doi: 10.1016/j.tics.2017.06.005

11 

SterzerP, SteinT, LudwigK, RothkirchM, HesselmannG. Neural processing of visual information under interocular suppression: A critical review. Front Psychol. 2014;5:453. doi: 10.3389/fpsyg.2014.00453

12 

ReesG. Neural correlates of the contents of visual awareness in humans. Philos Trans R Soc B Biol Sci. 2007;362:87786. doi: 10.1098/rstb.2007.2094

13 

KouiderS, DehaeneS. Levels of processing during non-conscious perception: A critical review of visual masking. Philos Trans R Soc B Biol Sci. 2007;362:85775. doi: 10.1098/rstb.2007.2093

14 

HassinRR. Yes it can: On the functional abilities of the human unconscious. Perspect Psychol Sci. 2013;8:195207. doi: 10.1177/1745691612460684

15 

SklarAY, LevyN, GolsteinA, MandelR, MarilA, HassinRR. Reading and doing arithmetic nonconsciously. Proc Natl Acad Sci U S A. 2012;109:196149. doi: 10.1073/pnas.1211645109

16 

HesselmannG. Dissecting visual awareness with fMRI. Neuroscientist. 2013;19:495508. doi: 10.1177/1073858413485988

17 

RamsøyTZ, OvergaardM. Introspection and subliminal perception. Phenomenol Cogn Sci. 2004;3:123.

18 

MaroisR, YiDJ, ChunMM. The neural fate of consciously perceived and missed events in the attentional blink. Neuron. 2004;41:46572. doi: 10.1016/s0896-6273(04)00012-1

19 

BeckDM, ReesG, FrithCD, LavieN. Neural correlates of change detection and change blindness. Nat Neurosci. 2001;4:64550. doi: 10.1038/88477

20 

KingJR, PescetelliN, DehaeneS. Brain mechanisms underlying the brief maintenance of seen and unseen sensory Information. Neuron. 2016;92:112234. doi: 10.1016/j.neuron.2016.10.051

21 

TrübutschekD, MartiS, OjedaA, KingJR, MiY, TsodyksM et al. A theory of working memory without consciousness or sustained activity. Elife. 2017;6:e23871. doi: 10.7554/eLife.23871

22 

TrübutschekD, MartiS, UeberschärH, DehaeneS. Probing the limits of activity-silent non-conscious working memory. Proc Natl Acad Sci U S A. 2019;116:1435867. doi: 10.1073/pnas.1820730116

23 

SotoD, MäntyläT, SilvantoJ. Working memory without consciousness. Curr Biol. 2011;21:R9123. doi: 10.1016/j.cub.2011.09.049

24 

BergströmF, ErikssonJ. The conjunction of non-consciously perceived object identity and spatial position can be retained during a visual short-term memory task. Front Psychol. 2015;6:1470. doi: 10.3389/fpsyg.2015.01470

25 

HesselmannG, HebartM, MalachR. Differential BOLD activity associated with subjective and objective reports during ‘blindsight’ in normal observers. J Neurosci. 2011;31:1293644. doi: 10.1523/JNEUROSCI.1556-11.2011

26 

SchmidtT. Invisible stimuli, implicit thresholds: Why invisibility judgments cannot be interpreted in isolation. Adv Cogn Psychol. 2015;11:3141. doi: 10.5709/acp-0169-3

27 

MacmillanNA, CreelmanCD. Detection theory: A user’s guide. Lawrence Erlbaum; 2005.

28 

BjörkmanM, JuslinP, WinmanA. Realism of confidence in sensory discrimination: The underconfidence phenomenon. Percept Psychophys. 1993;54:7581. doi: 10.3758/bf03206939

29 

PetersMAK, LauH. Human observers have optimal introspective access to perceptual processes even for visually masked stimuli. Elife. 2015;4:e09651. doi: 10.7554/eLife.09651

30 

EriksenCW. Discrimination and learning without awareness: A methodological survey and evaluation. Psychol Rev. 1960;67:279300. doi: 10.1037/h0041622

31 

HolenderD. Semantic activation without conscious identification in dichotic listening, parafoveal vision, and visual masking: A survey and appraisal. Behav Brain Sci. 1986;9:123.

32 

SchmidtT, VorbergD. Criteria for unconscious cognition: Three types of dissociation. Percept Psychophys. 2006;68:489504. doi: 10.3758/bf03193692

33 

SterzerP, HaynesJD, ReesG. Fine-scale activity patterns in high-level visual areas encode the category of invisible objects. J Vis. 2008;8:112. doi: 10.1167/8.15.10

34 

HaynesJD, ReesG. Predicting the orientation of invisible stimuli from activity in human primary visual cortex. Nat Neurosci. 2005;8:68691. doi: 10.1038/nn1445

35 

JiangY, ZhouK, HeS. Human visual cortex responds to invisible chromatic flicker. Nat Neurosci. 2007;10:65762. doi: 10.1038/nn1879

36 

JiangY, HeS. Cortical Responses to Invisible Faces: Dissociating Subsystems for Facial-Information Processing. Curr Biol. 2006;16:20239. doi: 10.1016/j.cub.2006.08.084

37 

MoutoussisK, ZekiS. The relationship between cortical activation and perception investigated with invisible stimuli. Proc Natl Acad Sci U S A. 2002;99:952732. doi: 10.1073/pnas.142305699

38 

FangF, HeS. Cortical responses to invisible objects in the human dorsal and ventral pathways. Nat Neurosci. 2005;8:13805. doi: 10.1038/nn1537

39 

FahrenfortJJ, SnijdersTM, HeinenK, van GaalS, ScholteHS, LammeVAF. Neuronal integration in visual cortex elevates face category tuning to conscious face perception. Proc Natl Acad Sci U S A. 2012;109:215049. doi: 10.1073/pnas.1207414110

40 

DehaeneS, ArtigesE, NaccacheL, MartelliC, ViardA, SchürhoffF et al. Conscious and subliminal conflicts in normal subjects and patients with schizophrenia: The role of the anterior cingulate. Proc Natl Acad Sci U S A. 2003;100:137227. doi: 10.1073/pnas.2235214100

41 

DehaeneS, NaccacheL, CohenL, Le BihanD, ManginJF, PolineJB et al. Cerebral mechanisms of word masking and unconscious repetition priming. Nat Neurosci. 2001;4:7528. doi: 10.1038/89551

42 

Van GaalS, RidderinkhofKR, ScholteHS, LammeVAF. Unconscious activation of the prefrontal no-go network. J Neurosci. 2010;30:414350. doi: 10.1523/JNEUROSCI.2992-09.2010

43 

ReingoldEM, MeriklePM. Using direct and indirect measures to study perception without awareness. Percept Psychophys. 1988;44:56375. doi: 10.3758/bf03207490

44 

MalachR, ReppasJB, BensonRR, KwongKK, JiangH, KennedyWA et al. Object-related activity revealed by functional magnetic resonance imaging in human occipital cortex. Proc Natl Acad Sci U S A. 1995;92:81359. doi: 10.1073/pnas.92.18.8135

45 

PourtoisG, SchwartzS, SpiridonM, MartuzziR, VuilleumierP. Object representations for multiple visual categories overlap in lateral occipital and medial fusiform cortex. Cereb Cortex. 2009;19:180619. doi: 10.1093/cercor/bhn210

46 

PitcherD, WalshV, DuchaineB. The role of the occipital face area in the cortical face perception network. Exp Brain Res. 2011;209:48193. doi: 10.1007/s00221-011-2579-1

47 

DilksDD, JulianJB, PaunovAM, KanwisherN. The occipital place area is causally and selectively involved in scene perception. J Neurosci. 2013;33:13316. doi: 10.1523/JNEUROSCI.4081-12.2013

48 

KanwisherN. Functional specificity in the human brain: A window into the functional architecture of the mind. Proc Natl Acad Sci U S A. 2010;107:1116370. doi: 10.1073/pnas.1005062107

49 

WagenmakersEJ, LoveJ., MarsmanM, JamilT, LyA, VerhagenJ et al. Bayesian inference for psychology. Part II: Example applications with JASP. Psychon Bull Rev. 2018;25:5876. doi: 10.3758/s13423-017-1323-7

50 

HesselmannG, DarcyN, LudwigK, SterzerP. Priming in a shape task but not in a category task under continuous flash suppression. J Vis. 2016;16:117.

51 

PeelenMV, Fei-FeiL, KastnerS. Neural mechanisms of rapid natural scene categorization in human visual cortex. Nature. 2009;460:947. doi: 10.1038/nature08103

52 

PeelenMV. KastnerS. A neural basis for real-world visual search in human occipitotemporal cortex. Proc Natl Acad Sci U S A. 2011;108:1212530. doi: 10.1073/pnas.1101042108

53 

KampsFS, JulianJB, KubiliusJ, KanwisherN, DilksDD. The occipital place area represents the local elements of scenes. Neuroimage. 2016;132:41724. doi: 10.1016/j.neuroimage.2016.02.062

54 

KriegeskorteN, GoebelR, BandettiniP. Information-based functional brain mapping. Proc Natl Acad Sci U S A. 2006;103:38638. doi: 10.1073/pnas.0600244103

55 

BenjaminiY, HochbergY. Controlling the false discovery rate: A practical and powerful approach to multiple testing. J R Stat Soc Ser B. 1995;57:289300.

56 

MumfordJA, PolineJ-B, PoldrackRA. Orthogonalization of regressors in fMRI models. PLoS ONE. 2015;10:e0126255. doi: 10.1371/journal.pone.0126255

57 

AzzopardiP. CoweyA. Is blindsight like normal, near-threshold vision? Proc Natl Acad Sci U S A. 1997;94:141904. doi: 10.1073/pnas.94.25.14190

58 

RodriguezV, ThompsonR, StokesM, BrettM, AlvarezI, Valdes-SosaM et al. Absence of face-specific cortical activity in the complete absence of awareness: Converging evidence from functional magnetic resonance imaging and event-Related potentials. J Cogn Neurosci. 2012;24:396415. doi: 10.1162/jocn_a_00137

59 

GayetS, GuggenmosM, ChristophelTB, HaynesJD, PaffenCLE, SterzerP et al. No evidence for mnemonic modulation of interocularly suppressed visual input. Neuroimage. 2020;215:116801. doi: 10.1016/j.neuroimage.2020.116801

60 

LudwigK, KathmannN, SterzerP, HesselmannG. Investigating category- and shape-selective neural processing in ventral and dorsal visual stream under interocular suppression. Hum Brain Mapp. 2015;36:13749. doi: 10.1002/hbm.22618

61 

DienesZ. How Bayesian statistics are needed to determine whether mental states are unconscious. In: OvergaardM, editor. Behavioral Methods in Consciousness Research. Oxford Academic; 2015. p. 199220.

62 

ShanksDR. Regressive research: The pitfalls of post hoc data selection in the study of unconscious mental processes. Psychon Bull Rev. 2017;24:75275. doi: 10.3758/s13423-016-1170-y

63 

SteinT, UtzV, van OpstalF. Unconscious semantic priming from pictures under backward masking and continuous flash suppression. Conscious Cogn. 2020;78:102864. doi: 10.1016/j.concog.2019.102864

64 

Van Den BusscheE, NotebaertK, ReynvoetB. Masked primes can be genuinely semantically processed: A picture prime study. Exp Psychol. 2009;56:295300. doi: 10.1027/1618-3169.56.5.295

65 

VorbergD, MattlerU, HeineckeA, SchmidtT, SchwarzbachJ. Different time courses for visual perception and action priming. Proc Natl Acad Sci U S A. 2003;100:627580. doi: 10.1073/pnas.0931489100

66 

KlotzW, NeumannO. Motor activation without conscious discrimination in metacontrast masking. J Exp Psychol Hum Percept Perform. 1999;25:97692.

67 

CheesmanJ, MeriklePM. Priming with and without awareness. Percept Psychophys. 1984;36:38795. doi: 10.3758/bf03202793

68 

ScottRB, SamahaJ, ChrisleyR, DienesZ. Prevailing theories of consciousness are challenged by novel cross-modal associations acquired between subliminal stimuli. Cognition. 2018;175:16985. doi: 10.1016/j.cognition.2018.02.008

69 

SotoD, SilvantoJ. Reappraising the relationship between working memory and conscious awareness. Trends Cogn Sci. 2014;18:5205. doi: 10.1016/j.tics.2014.06.005

70 

CharlesL, Van OpstalF, MartiS, DehaeneS. Distinct brain mechanisms for conscious versus subliminal error detection. Neuroimage. 2013;73:8094. doi: 10.1016/j.neuroimage.2013.01.054

71 

SaltiM, MonfoS, CharlesL, KingJR, ParkkonenL, DehaeneS Distinct cortical codes and temporal dynamics for conscious and unconscious percepts. Elife. 2015;4:e05652. doi: 10.7554/eLife.05652

72 

ReberTP, FaberJ, NiediekJ, BoströmJ, ElgerCE, MormannF Single-neuron correlates of conscious perception in the human medial temporal lobe. Curr Biol. 2017;27, e2:2991–8. doi: 10.1016/j.cub.2017.08.025

73 

SchurgerA, SarigiannidisI, NaccacheL, SittJD, DehaeneS. Cortical activity is more stable when sensory stimuli are consciously perceived. Proc Natl Acad Sci U S A. 2015;112:E208392. doi: 10.1073/pnas.1418730112

74 

WatanabeM, ChengK, MurayamaY, UenoK, AsamizuyaT, TanakaK et al. Attention but not awareness modulates the BOLD signal in the human V1 during binocular suppression. Science. 2011;334:82931. doi: 10.1126/science.1203161

75 

SteinT, PeelenMV, SterzerP. Eye gaze adaptation under interocular suppression. J Vis. 2012;12:117. doi: 10.1167/12.7.1

76 

MelloniL, SchwiedrzikCM, MüllerN, RodriguezE, SingerW. Expectations change the signatures and timing of electrophysiological correlates of perceptual awareness. J Neurosci. 2011;31:138696. doi: 10.1523/JNEUROSCI.4570-10.2011

77 

ManiscalcoB. LauH. A signal detection theoretic approach for estimating metacognitive sensitivity from confidence ratings. Conscious Cogn. 2012;21:42230. doi: 10.1016/j.concog.2011.09.021

78 

ZekiS. A Theory of Micro-Consciousness. In: The Blackwell Companion to Consciousness. Blackwell Publishing; 2007. p. 580588.

79 

MalachR. The measurement problem in consciousness research. Behav Brain Sci. 2007;30:5167.

80 

KochC, MassiminiM, BolyM, TononiG. Neural correlates of consciousness: Progress and problems. Nat Rev Neurosci. 2016;17:30721. doi: 10.1038/nrn.2016.22

81 

MashourGA, RoelfsemaP, ChangeuxJP, DehaeneS. Conscious processing and the global neuronal workspace hypothesis. Neuron. 2020;105:77698. doi: 10.1016/j.neuron.2020.01.026

82 

TsuchiyaN, WilkeM, FrässleS, LammeVAF. No-report paradigms: Extracting the true neural correlates of consciousness. Trends Cogn Sci. 2015;19:75770. doi: 10.1016/j.tics.2015.10.002

83 

ButtonKS, IoannidisJPA, MokryszC, NosekBA, FlintJ, RobinsonESJ et al. Power failure: Why small sample size undermines the reliability of neuroscience. Nat Rev Neurosci. 2013;14:36576. doi: 10.1038/nrn3475

84 

VadilloMA, LinssenD, OrgazC, ParsonsS, ShanksDR. Unconscious or underpowered? Probabilistic cuing of visual attention. J Exp Psychol Gen. 2020;149:16081. doi: 10.1037/xge0000632

85 

VadilloMA, KonstantinidisE, ShanksDR. Underpowered samples, false negatives, and unconscious learning. Psychon Bull Rev. 2016;23:87102. doi: 10.3758/s13423-015-0892-6

86 

BrainardDH. The psychophysics toolbox. Spat Vis. 1997;10:4336.

87 

EbnerNC, RiedigerM. LindenbergerU. FACES–A database of facial expressions in young, middle-aged, and older women and men: Development and validation. Behav Res Methods. 2010;42:35162. doi: 10.3758/BRM.42.1.351

88 

EstebanO, MarkiewiczCJ, BlairRW, MoodieCA, IsikAI, ErramuzpeA et al. fMRIPrep: A robust preprocessing pipeline for functional MRI. Nat Methods. 2019;16:1116. doi: 10.1038/s41592-018-0235-4

89 

FonovV, EvansA, McKinstryR, AlmliC, CollinsD. Unbiased nonlinear average age-appropriate brain templates from birth to adulthood. Neuroimage. 2009;47:S102.

90 

OosterhofNN, ConnollyAC, HaxbyJV. CoSMoMVPA: Multi-modal multivariate pattern analysis of neuroimaging data in matlab/GNU octave. Front Neuroinform. 2016;10:1027. doi: 10.3389/fninf.2016.00010

91 

WangL, MruczekREB, ArcaroMJ, KastnerS. Probabilistic maps of visual topography in human cortex. Cereb Cortex. 2015;25:391131. doi: 10.1093/cercor/bhu277

92 

JulianJB, FedorenkoE, WebsterJ, KanwisherN. An algorithmic method for functionally defining regions of interest in the ventral visual pathway. Neuroimage. 2012;60:235764. doi: 10.1016/j.neuroimage.2012.02.055

93 

JASP Team. JASP (Version 0.12.2). Comput Softw. 2020.

25 Nov 2020

Dear Dr Stein,

Thank you for submitting your manuscript entitled "Processing of subjectively and objectively invisible stimuli in human visual cortex" for consideration as a Research Article by PLOS Biology.

Your manuscript has now been evaluated by the PLOS Biology editorial staff, as well as by an academic editor with relevant expertise, and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by Dec 01 2020 11:59PM. (Note the office will be close Thursday 26 and Friday 27 of November).

Login to Editorial Manager here: https://www.editorialmanager.com/pbiology

During resubmission, you will be invited to opt-in to posting your pre-review manuscript as a bioRxiv preprint. Visit http://journals.plos.org/plosbiology/s/preprints for full details. If you consent to posting your current manuscript as a preprint, please upload a single Preprint PDF when you re-submit.

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Given the disruptions resulting from the ongoing COVID-19 pandemic, please expect delays in the editorial process. We apologise in advance for any inconvenience caused and will do our best to minimize impact as far as possible.

Feel free to email us at plosbiology@plos.org if you have any queries relating to your submission.

Kind regards,

Gabriel Gasque, Ph.D.,

Senior Editor

PLOS Biology


3 Feb 2021

Dear Dr Stein,

Thank you very much for submitting your manuscript "Processing of subjectively and objectively invisible stimuli in human visual cortex" for consideration as a Research Article at PLOS Biology. Your manuscript has been evaluated by the PLOS Biology editors, by an academic editor with relevant expertise, and by two independent reviewers. Please accept my apologies for the delay in sending the decision below to you.

In light of the reviews and the academic editor's detailed comments (below), we are pleased to offer you the opportunity to address the comments from the academic editor and from the reviewers in a revised version that we anticipate should not take you very long. We will then assess your revised manuscript and your response to the academic editor and reviewers' comments and we may consult the reviewers again.

We expect to receive your revised manuscript within 1 month.

Please email us (plosbiology@plos.org) if you have any questions or concerns, or would like to request an extension. At this stage, your manuscript remains formally under active consideration at our journal; please notify us by email if you do not intend to submit a revision so that we may end consideration of the manuscript at PLOS Biology.

**IMPORTANT - SUBMITTING YOUR REVISION**

Your revisions should address the specific points made by the academic editor and by each reviewer. As you will see from their detailed comments, the academic editor and the two reviewers feel that this topic is important and timely, and note the careful study design. There was, however, some concerns (raised most strongly by reviewer 3) with the way in which objective and subjective invisibility was defined and achieved - that reviewers argue could affect the study interpretations. Please address these comments thoroughly.

Please also make sure to address the data and other policy-related requests noted at the end of this email.

Please submit the following files along with your revised manuscript:

1. A 'Response to Reviewers' file - this should detail your responses to the editorial requests, present a point-by-point response to all of the reviewers' comments, and indicate the changes made to the manuscript.

*NOTE: In your point by point response to the reviewers, please provide the full context of each review. Do not selectively quote paragraphs or sentences to reply to. The entire set of reviewer comments should be present in full and each specific point should be responded to individually.

You should also cite any additional relevant literature that has been published since the original submission and mention any additional citations in your response.

2. In addition to a clean copy of the manuscript, please also upload a 'track-changes' version of your manuscript that specifies the edits made. This should be uploaded as a "Related" file type.

*Resubmission Checklist*

When you are ready to resubmit your revised manuscript, please refer to this resubmission checklist: https://plos.io/Biology_Checklist

To submit a revised version of your manuscript, please go to https://www.editorialmanager.com/pbiology/ and log in as an Author. Click the link labelled 'Submissions Needing Revision' where you will find your submission record.

Please make sure to read the following important policies and guidelines while preparing your revision:

*Published Peer Review*

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out. Please see here for more details:

https://blogs.plos.org/plos/2019/05/plos-journals-now-open-for-published-peer-review/

*PLOS Data Policy*

Please note that as a condition of publication PLOS' data policy (http://journals.plos.org/plosbiology/s/data-availability) requires that you make available all data used to draw the conclusions arrived at in your manuscript. If you have not already done so, you must include any data used in your manuscript either in appropriate repositories, within the body of the manuscript, or as supporting information (N.B. this includes any numerical values that were used to generate graphs, histograms etc.). For an example see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5

*Blot and Gel Data Policy*

We require the original, uncropped and minimally adjusted images supporting all blot and gel results reported in an article's figures or Supporting Information files. We will require these files before a manuscript can be accepted so please prepare them now, if you have not already uploaded them. Please carefully read our guidelines for how to prepare and upload this data: https://journals.plos.org/plosbiology/s/figures#loc-blot-and-gel-reporting-requirements

*Protocols deposition*

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosbiology/s/submission-guidelines#loc-materials-and-methods

Thank you again for your submission to our journal. We hope that our editorial process has been constructive thus far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Gabriel Gasque, Ph.D.,

Senior Editor,

ggasque@plos.org,

PLOS Biology

*****************************************************

ETHICS:

-- Please indicate within your manuscript if your experimental procedures adhered to principles expressed in the Declaration of Helsinki or any other national or international ethical guidelines.

*****************************************************

DATA:

You may be aware of the PLOS Data Policy, which requires that all data be made available without restriction: http://journals.plos.org/plosbiology/s/data-availability. For more information, please also see this editorial: http://dx.doi.org/10.1371/journal.pbio.1001797

I note that you have stated your data in confidential. Please refer to these guidelines for accepted data access restrictions: https://journals.plos.org/plosbiology/s/data-availability#loc-acceptable-data-access-restrictions

Note, nonetheless, that we do not require the raw data. Rather, we ask for all individual quantitative observations that underlie the data summarized in the figures and results of your paper. For an example see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5

These data can be made available in one of the following forms:

1) Supplementary files (e.g., excel). Please ensure that all data files are uploaded as 'Supporting Information' and are invariably referred to (in the manuscript, figure legends, and the Description field when uploading your files) using the following format verbatim: S1 Data, S2 Data, etc. Multiple panels of a single or even several figures can be included as multiple sheets in one excel file that is saved using exactly the following convention: S1_Data.xlsx (using an underscore).

2) Deposition in a publicly available repository. Please also provide the accession code or a reviewer link so that we may view your data before publication.

Regardless of the method selected, please ensure that you provide the individual numerical values that underlie the summary data displayed in the following figure panels: Figures 1b-e, 2ab, 3ab, 4ab, 5a-c, S1bc, S2ab, S3ab, S4ab, and S5 (Please note that there are two figures S3, I thus renumbered S3 “(a) Results from the additional searchlight analyses of the main experiment only, …” as S4, and renumbered S4 as S5).

NOTE: the numerical data provided should include all replicates AND the way in which the plotted mean and errors were derived (it should not present only the mean/average values).

Please also ensure that each figure legend in your manuscript includes information on where the underlying data can be found and that your supplemental data file/s has/have a legend.

Please ensure that your Data Statement in the submission system accurately describes where your data can be found.

*****************************************************

REVIEWS:

Academic Editor: The paper by Stein et al addresses a fundamental question in consciousness research- to what extent invisible stimuli nevertheless activate in an informative manner the human visual system. This is an important issue- since it bears critical importance to any theory that attempts to delineate the neuronal mechanisms that underlie conscious vs. non-conscious visual processes.

While this question has been addressed by a number of previous studies- these studies were notoriously plagued by ungrounded assumptions and poor methodology. Stein et al has made, to my mind, an important and decisive advance in this important field by carefully noting these previous problems and designing their study in a rigorous manner that impressively addresses these problems at their core. In that sense I feel that Stein et al bring a breath of fresh air to this problematic field- and provide conclusive and convincing findings in which the confounds and methodological mistakes are cleared out. I feel strongly that this work will provide an important and significant advance in this major field of human vision research.

Below are my comments following the flow of the manuscript itself.

Abstract:

"We show that neural representations of objectively invisible faces and houses

are limited to visual (shape-related) object properties, while subjectively invisible stimuli

are processed up to more abstract, categorical levels of representation"- I think the contrast between visual and abstract is misleading- it implies that category-selective areas are not visual- which is unfounded. The authors should use more neutral terms- such as early or lower-level vs. higher level visual areas.

Results:

Figure 1:

It will be informative if the authors show the fit between panels b and c- by normalizing the values between 0 and 100% and superimposing the two curves. The fit should be pointed out, as well as the highly non-linear (sigmoid) nature of the dependence on mask contrast.

Discussion:

General points:

It is important that the authors discuss the meaning of the "stronger representation" to use their terminology when analyzed through multivariate correlation patterns. Specifically- a stronger vs weaker representation may result from at least two main sources: first, a weaker univariate activation- i.e. lower signal to noise responses in each voxel. Second- it may result from a change in the pattern of activation elicited during the weak vs strong responses-resulting in weaker and stronger similarity to the bench mark patterns. This point needs to be clarified- and kept in mind when interpreting the results. Just as an example is the following conclusion: "The present findings will resonate with the idea that subjectively invisible stimuli are processed in a way that is qualitatively similar to (clearly) visible stimuli"- In this statement the authors, presumably, assume that the patterns of activation for the subjectively invisible and visible stimuli are similar-differing only in signal amplitude. However this conclusion ignores the fact that the massive reduction in representation strength for subjectively invisible stimuli may be the result of a corresponding change in the pattern of activation making it dissimilar to the benchmark pattern.

A second major methodological issue that should be discussed is the critical difference in the response properties to visible stimuli in early vs high order cortex- specifically in this case to the mask vs. target stimuli. It is quite likely that the weak category representations they find in V1 during the visible conditions is due to literally masking the category information by the much stronger mask response- since the fMRI method is too sluggish to temporally separate out the target response from the far larger response to the masks. This can easily lead to an erroneous underestimation of the low-level differences between the categories in V1 even for visible stimuli. This point should be addressed by the authors.

Specific points:

Before jumping into the discussion of the brain activations associated with the invisible conditions- it is important that the authors will first note the main, and very striking, effects demonstrated by their study. In particular-the very dramatic change in the "visibility effect"- i.e. the difference between visible vs invisible conditions- as one moves from early (essentially no difference) to high order visual cortex (massive difference). Similarly they should point in this context the abolishment of such visibility effect in frontal cortex.

"Rendering stimuli perceptually indiscriminable requires very low stimulus strength" This is incorrect- there are cases where strong stimuli can be rendered invisible - such as binocular rivalry, crowding and motion-induced blindness.

"above-chance performance for subjectively invisible stimuli simply means that

these stimuli were in fact not invisible"-

A possible analysis that may help clarify this conundrum could be to analyze separately those trials in which participants marked the target as invisible but guessed the category correctly vs. those subjectively invisible trials in which the participants guessed incorrectly.

"objectively invisible stimuli can be decoded from patterns of fMRI activity in human visual cortex- to be fair to their own data the authors should add "albeit at a drastically lower strength than visible stimuli"

Reviewer #2: This is a very interesting and carefully conducted study. The methods and results are solid. I have no issues with the design or analyses. My only concerns are with the interpretation of the results and the way in which the conclusions and implications are framed in the abstract vs. discussion.

1. On page 14 (middle of first full paragraph), the authors state, "Category information could be decoded from early visual cortex, from object-selective LOC, from the (largely overlapping) category-selective areas in lateral occipital cortex (OFA/OPA), and, less robustly, from ventrotemporal areas (FFA/PPA), while participants were unable to discriminate these categories, thereby satisfying the most rigorous criteria for establishing absence of awareness"

I don't quite understand how this conclusion is consistent with the general point of the paper as outlined in the abstract. To render stimuli objectively invisible, a stronger mask was used, and this stronger mask led to weaker stimulus-evoked activity patterns in all of the brain regions analyzed. But, even with objective invisibility, the authors still find above chance decoding in all of these regions, so doesn't this mean that even with stricter criteria and improved methods (objective instead of subjective invisibility), the extent of unconscious visual processing is quite vast? Perhaps I'm not understanding the authors' interpretation of (A) the existence of significant multi-voxel patterns vs. (B) magnitude differences between subj vs. obj invisibility. (A) seems more relevant to the general question of unconscious information processing while (B) seems more relevant to a methodological nuance. Indeed, on the following page, the authors state, "Notwithstanding these limitations, our findings provide more conclusive fMRI evidence for unconscious processing of objectively invisible stimuli in human visual cortex."

My general concern here is that what is stated in the abstract doesn't really match up with this main conclusion. The abstract states, "These results demonstrate that the hypothesized extent of unconscious information processing is determined by the measurement approach. Furthermore, our data show that subjective and objective approaches are associated with different neural correlates of consciousness and thus have implications for neural theories of consciousness." But one of the main conclusions seems to be that decoding was still possible in a ton of areas (EVC, LOC, OPA, OFA, PPA, FFA), even for objectively invisible stimuli.

How does this finding match-up with the claims towards the end of the abstract? What are the implications for neural theories of consciousness, and how do these implications differ for obj-vis vs. obj-inv contrasts compared to subj-vis vs. subj-inv contrasts? The key question here is whether the neural results just change in magnitude or in kind when these different measurements of awareness are used.

2. Why does figure 3b not include obj-vis vs. obj-inv? A major focus of this paper is about how objective versus subjective measures of awareness lead to different neural results. Why only show subj-vis vs. subj-inv and subj-inv vs. obj-inv? Obj-vis vs. obj-inv seems just as interesting of a contrast for the purposes of this study.

3. On p. 8, first full paragraph, the authors state, "...participants did a simple one-back task on stimuli presented in a block design. This approach ... is often considered to isolate perceptual representations from later cognitive processes." I don't think this is the case. A one-back task requires subjects to hold each image in memory to see if it matches the next image. Holding things in memory seems like a "later cognitive process" and it's unclear how such a task isolates perceptual representations from memory processes.

4. Small but important typo on p. 5, second to last sentence in first paragraph of results: "(subj-vis: mean contrast 7.8%, SD 4.4; subj-vis: mean contrast 11.9%, SD 4.3)." I believe the first label should be "subj-inv".

Reviewer #3: This study is an important and timely attempt to compare neural activations evoked by stimuli that are rendered invisible, and that are either subjectively or objectively invisible. I absolutely love the research question and could not be more enthusiastic about it and about the importance of this study. I am no fMRI expert, but to me the analyses seemed compelling and comprehensive, and the introduction and discussion were well built, interesting, and provided a good description of the literature. The study also included a large sample size, which is highly important, especially for the small effects that are typical of the field. So my overall impression of this study is very positive.

However, and unfortunately given my overall enthusiasm about this manuscript, I am concerned that the way objective and subjective invisibility were defined are not optimal, and increase the chances of finding differences between objective and subjective invisibility, as they lower the signal for the objective case and allow a stronger signal (or contamination by conscious perception) for the subjective case. Thus, the reported differences might not be an accurate estimation of the actual differences between objective and subjective invisibility, when operationalized correctly (see below). I would have loved to see a study where objective invisibility was determined per individual, and subjective invisibility would have indeed been defined as the lowest PAS level (which the authors claim is the case, but I disagree, as I explain below). I cannot understate how much I think such a study is important. In the absence of such a clean comparison, I would advise that the authors substantially tone down the claims about the differences between the two types of invisibility, acknowledge the limitations of their design, and instead emphasize the finding of unconscious processing even under the strict conditions of objective invisibility - a finding that is highly important (see point 3 below). To summarize, I find the study important and interesting, and would urge the authors to frame their results differently given the concerns I've raised.

Major comments:

1. My first concern relates to the objective condition - typically, experimenters try to reach the sweet spot where the physical signal is as strong as possible, while objective performance is at chance. Although this was done here, to some extent, in the pilot experiment probing masking efficiency, it was not done well enough, I am afraid. The authors chose the maximal contrast of the mask based on the group performance in the pilot study Could it be that this manipulation is too strong, and substantially abolishes the signal evoked by the stimuli? A much preferable approach, which is often taken by researchers (e.g., Hesselmann et al., 2006, JoV) is to first calibrate the physical parameters (e.g., contrast) per subject, given the known differences between individuals, and find that sweet spot for each one. This is much more likely to allow higher-level processing also using the objective threshold. To take the current approach ad absurdum, one could also simply present the critical stimuli at 0% contrast, get objective chance performance and no signal whatsoever… This is of course not the case here (and indeed effects were found), but the authors should still explain why they chose not to calibrate per participant to allow for the strongest possible signal in the objective condition, and acknowledge that this could be the cause for the difference they find between the two conditions.

2. The above issue is even more problematic when taking into account the definition of subjective visibility, which acts in the opposite direction; if the objective threshold is too strict, the subjective one seems too lenient. The authors write that subjects were instructed to report the stimuli as visible "even when they had only a vague idea of the stimulus category", and accordingly, claim that "subjective invisibility on our binary scale corresponded to the lowest visibility level of more fine-grained visibility scales, such as the four-point Perceptual Awareness Scale". I am not convinced that this is accurate, as the second level in PAS is "A feeling that something has been shown. Not characterised by any content, and this cannot be specified any further" (Ramsøy Overgaard, 2004). Thus, having a vague idea of the stimulus category is actually more visible than the second level, where one should have *no* access to any content about the stimulus - they only know something was there, but don't have any knowledge about what it was, not even a vague idea. This suggests that the "invisible" rating in this study actually reflects visibility levels 1 and 2, and not only 1. Unfortunately, this does pose a substantial problem, as the entire claim made here is based on the difference between subjective and objective threshold. But this comparison is now more problematic, as we are no longer comparing the lowest subjective invisibility with objective invisibility. I am not sure how this can be solved. To the very least, the authors should acknowledge this limitation and change the claims so to refer to the two lowest levels of subjective invisibility. The latter have been used in some studies to define invisibility (e.g., Melloni et al., 2011, J Neuroscience), but I wouldn't say it is the common practice in the field - typically, only the first level is used. Importantly, in many of these studies subjects are also at chance for the objective measures for visibility 1 trials - which is not the case here (and this might be explained by the fact that their "invisible" condition is not really invisible). Thus, I am not sure the critical test presented here - comparing the level of processing between objective and subjective invisibility - is a fair one.

3. I agree with the authors that this work is probably the most robust demonstration of unconscious processing in fMRI to date. Given the concerns I've raised above, I would emphasize this point more, and deemphasize the differences between the objective and the subjective conditions (as again, I think the comparison here is biased towards finding differences). The unconscious findings by themselves are highly interesting and important, and also set the bar for future studies aiming to make similar claims.

4. Was any correction for multiple comparisons across the paper? Such a correction is in order (e.g., FDR). The latter seemed to have been applied on the searchlight analysis, but there were many others conducted throughout the manuscript that should also be corrected for (for an explanation about the importance of correcting comparisons throughout the paper, see Benjamini et al., 2001, Behavioral Brain Research).

5. When talking about objective measures potentially underestimating unconscious processing, the authors should also mention as one of the reasons that above-chance performance in the objective task might be driven by unconscious effects.

6. The authors rightfully note that their sample size is larger than the typical fMRI studies in the field. But how was it determined? Was there any rationale behind it?

Minor comments:

1. I find it somewhat strange that when talking about "all major theories of consciousness", IIT is not mentioned. If the authors think it does not assume unconscious processing (which I don't believe is the case), they should change the text to "most major theories of consciousness". Otherwise, I think it should be cited there as well.

2. I apologize but I don't understand what is depicted in the line plots in figure 2. Are these simply lines connecting between the "peaks" of the bar plots? I don't really see why they are needed, as the readers can simply imagine them when looking at the bars (they are also somewhat misleading, as the variable is not continuous). I would remove them (or, if the authors really think they are necessary, add them to the bars plot. But I would advise to remove). If this is not what they depict, then please explain what is represented there.


12 Mar 2021


19 Apr 2021

Dear Dr Stein,

Thank you for submitting your revised Research Article entitled "Processing of subjectively and objectively invisible stimuli in human visual cortex" for publication in PLOS Biology. I have now obtained advice from the original reviewers and have discussed their comments with the Academic Editor. 

Based on the reviews, we will probably accept this manuscript for publication, provided you satisfactorily address the remaining points raised by reviewer 3. Please also make sure to address the data and other policy-related requests listed below my signature.

Editorially, we would like to make your title a bit more instructive, rather than descriptive, and suggest using: “The human visual system differentially represents subjectively and objectively invisible stimuli.” We would be happy to work with on a similar alternative.

As you address these items, please take this last chance to review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the cover letter that accompanies your revised manuscript.

We expect to receive your revised manuscript within two weeks.

To submit your revision, please go to https://www.editorialmanager.com/pbiology/ and log in as an Author. Click the link labelled 'Submissions Needing Revision' to find your submission record. Your revised submission must include the following:

-  a cover letter that should detail your responses to any editorial requests, if applicable, and whether changes have been made to the reference list

-  a Response to Reviewers file that provides a detailed response to the reviewers' comments (if applicable)

-  a track-changes file indicating any changes that you have made to the manuscript. 

NOTE: If Supporting Information files are included with your article, note that these are not copyedited and will be published as they are submitted. Please ensure that these files are legible and of high quality (at least 300 dpi) in an easily accessible file format. For this reason, please be aware that any references listed in an SI file will not be indexed. For more information, see our Supporting Information guidelines:

https://journals.plos.org/plosbiology/s/supporting-information  

*Published Peer Review History*

Please note that you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out. Please see here for more details:

https://blogs.plos.org/plos/2019/05/plos-journals-now-open-for-published-peer-review/

*Early Version*

Please note that an uncorrected proof of your manuscript will be published online ahead of the final version, unless you opted out when submitting your manuscript. If, for any reason, you do not want an earlier version of your manuscript published online, uncheck the box. Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us as soon as possible if you or your institution is planning to press release the article.

*Protocols deposition*

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please do not hesitate to contact me should you have any questions.

Sincerely,

Gabriel Gasque, Ph.D.,

Senior Editor,

ggasque@plos.org,

PLOS Biology

------------------------------------------------------------------------

ETHICS STATEMENT:

-- Please indicate within your manuscript is your experiments were conducted according to the principles expressed in the Declaration of Helsinki or any other national or international ethical guideline.

------------------------------------------------------------------------

DATA POLICY:

-- Thank you for uploading your quantitative data to OSF. I could not find files for Fig S1, even when the figure legend indicates the data should be there. Could you please either explain or upload those data as well? Please disregard this point if the data were already uploaded but I missed them.

------------------------------------------------------------------------

DATA NOT SHOWN?

- Please note that per journal policy, we do not allow the mention of "data not shown", "personal communication", "manuscript in preparation" or other references to data that is not publicly available or contained within this manuscript. Please either remove mention of these data or provide figures presenting the results and the data underlying the figure(s).

------------------------------------------------------------------------

Reviewer remarks:

Reviewer #2: The authors have addressed all of my concerns (and in my opinion, the other reviewers' concerns as well). I am happy to support a positive decision on this version of the manuscript.

Reviewer #3: The revised manuscript addresses all my concerns, and I would be happy to see it published. Here are a few minor suggestions which might strengthen the manuscript, in my opinion, and make it more accurate:

1. The running title is "subjective vs. objective measures of awareness", which in my opinion underrepresents the scope of the paper. I would change to: "processing subjectively vs. objectively invisible stimuli".

2. The addition of a disclaimer about the lack of individual calibration is excellent, but I wonder if readers who are not experts in the field would understand the key message. I accordingly suggest adding an explanation at the end. After "although we cannot exclude the possibility that our group-based calibration approach did not result in optimal stimulus strength for every individual observer", I would add: "This might have reduced the chances for finding higher-level processing also for objective-invisibility", or "This might have inflated the difference between objective invisibility and subjective invisibility in our study".

3. The figures were of low quality; I imagine that this is just some technical limitation of the review process, but write this warning to make sure this would not be the case for the published paper.


20 Apr 2021


20 Apr 2021

Dear Dr Stein,

On behalf of my colleagues and the Academic Editor, Rafael Malach, I am pleased to say that we can in principle offer to publish your Research Article "The human visual system differentially represents subjectively and objectively invisible stimuli" in PLOS Biology, provided you address any remaining formatting and reporting issues. These will be detailed in an email that will follow this letter and that you will usually receive within 2-3 business days, during which time no action is required from you. Please note that we will not be able to formally accept your manuscript and schedule it for publication until you have made the required changes.

***IMPORTANT: Please also make sure to add, within your manscript, an approval/ID number for your protocol approved by the University of Amsterdam ethics committee. (Please accept my apologies; I should have requested this information before).

Please take a minute to log into Editorial Manager at http://www.editorialmanager.com/pbiology/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with biologypress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

Thank you again for supporting Open Access publishing. We look forward to publishing your paper in PLOS Biology. 

Sincerely, 

Gabriel Gasque, Ph.D. 

Senior Editor 

PLOS Biology

https://creativecommons.org/licenses/by/4.0/This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
https://www.researchpad.co/tools/openurl?pubtype=article&doi=10.1371/journal.pbio.3001241&title=The human visual system differentially represents subjectively and objectively invisible stimuli&author=Timo Stein,Daniel Kaiser,Johannes J. Fahrenfort,Simon van Gaal,Rafael Malach,Gabriel Gasque,&keyword=&subject=Research Article,Biology and Life Sciences,Anatomy,Brain,Visual Cortex,Medicine and Health Sciences,Anatomy,Brain,Visual Cortex,Biology and Life Sciences,Neuroscience,Brain Mapping,Functional Magnetic Resonance Imaging,Medicine and Health Sciences,Diagnostic Medicine,Diagnostic Radiology,Magnetic Resonance Imaging,Functional Magnetic Resonance Imaging,Research and Analysis Methods,Imaging Techniques,Diagnostic Radiology,Magnetic Resonance Imaging,Functional Magnetic Resonance Imaging,Medicine and Health Sciences,Radiology and Imaging,Diagnostic Radiology,Magnetic Resonance Imaging,Functional Magnetic Resonance Imaging,Research and Analysis Methods,Imaging Techniques,Neuroimaging,Functional Magnetic Resonance Imaging,Biology and Life Sciences,Neuroscience,Neuroimaging,Functional Magnetic Resonance Imaging,Biology and Life Sciences,Neuroscience,Cognitive Science,Cognitive Psychology,Perception,Sensory Perception,Vision,Biology and Life Sciences,Psychology,Cognitive Psychology,Perception,Sensory Perception,Vision,Social Sciences,Psychology,Cognitive Psychology,Perception,Sensory Perception,Vision,Biology and Life Sciences,Neuroscience,Sensory Perception,Vision,Biology and Life Sciences,Neuroscience,Cognitive Science,Cognition,Biology and Life Sciences,Neuroscience,Cognitive Science,Cognitive Psychology,Perception,Sensory Perception,Biology and Life Sciences,Psychology,Cognitive Psychology,Perception,Sensory Perception,Social Sciences,Psychology,Cognitive Psychology,Perception,Sensory Perception,Biology and Life Sciences,Neuroscience,Sensory Perception,Biology and Life Sciences,Neuroscience,Cognitive Science,Cognitive Neuroscience,Consciousness,Biology and Life Sciences,Neuroscience,Cognitive Neuroscience,Consciousness,Biology and Life Sciences,Neuroscience,Cognitive Science,Cognitive Neuroscience,Consciousness,Theories of Consciousness,Biology and Life Sciences,Neuroscience,Cognitive Neuroscience,Consciousness,Theories of Consciousness,Biology and Life Sciences,Psychology,Behavior,Social Sciences,Psychology,Behavior,