RE S EARCH REPORT
Subliminal perception is continuous with conscious vision
and can be predicted from prestimulus
electroencephalographic activity
Henry Railo1,2,3 | Roberto Piccin4 | Karolina M. Lukasik5
1Department of Clinical Neurophysiology,
University of Turku, Turku, Finland
2Turku Brain and Mind Centre,
University of Turku, Turku, Finland
3Department of Psychology, University of
Turku, Turku, Finland
4Department of Life Sciences, University
of Trieste, Trieste, Italy
5Department of Psychology, Åbo Akademi
University, Turku, Finland
Correspondence
Henry Railo, Department of Clinical
Neurophysiology, University of Turku and
Turku University Hospital, Turku,
Finland.
Email: henry.railo@utu.fi
Funding information
Academy of Finland, Grant/Award
Number: 308533
Edited by: Niko Busch
Abstract
Individuals are able to discriminate visual stimuli they report not consciously
seeing. This phenomenon is known as “subliminal perception.” Such capacity is
often assumed to be relatively automatic in nature and rely on stimulus-driven
activity in low-level cortical areas. Instead, here we asked to what extent neural
activity before stimulus presentation influences subliminal perception. We asked
participants to discriminate the location of a briefly presented low-contrast
visual stimulus and then rate how well they saw the stimulus. Consistent with
previous studies, participants correctly discriminated with slightly above
chance-level accuracy the location of a stimulus they reported not seeing. Signal
detection analyses indicated that while subjects categorized their percepts as
“unconscious,” their capacity to discriminate these stimuli lay on the same
continuum as conscious vision. We show that the accuracy of discriminating the
location of a subliminal stimulus could be predicted with relatively high accu-
racy (AUC = 0.70) based on lateralized electroencephalographic (EEG) activity
before the stimulus, the hemifield where the stimulus was presented, and the
accuracy of previous trial’s discrimination response. Altogether, our results
suggest that rather than being a separate unconscious capacity, subliminal per-
ception is based on similar processes as conscious vision.
KEYWORD S
blindsight, consciousness, cortical excitability, EEG, oscillations, prestimulus, subliminal
perception, unconscious perception
1 | INTRODUCTION
One of the major aims of neuroscience is to uncover
which neural processes enable a person to consciously
experience and respond to stimuli in their environment.
Interestingly, visual stimuli that individuals report as not
consciously perceived can sometimes influence their
behavior, a phenomenon known as subliminal or
Abbreviations: AUC, area under the curve; BIC, Bayesian information criterion; EEG, electroencephalography; ERP, event-related potential; ERSP,
event-related spectral perturbation; VAN, visual awareness negativity.
Received: 1 July 2020 Revised: 8 June 2021 Accepted: 9 June 2021
DOI: 10.1111/ejn.15354
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2021 The Authors. European Journal of Neuroscience published by Federation of European Neuroscience Societies and John Wiley & Sons Ltd.
Eur J Neurosci. 2021;54:4985–4999. wileyonlinelibrary.com/journal/ejn 4985
unconscious perception (Hannula et al., 2005; Kouider &
Dehaene, 2007; Phillips, 2018; Stein & Peelen, 2021). The
neural mechanisms that enable this subliminal capacity
are not well understood, and debate continues on
whether subliminal perception reveals a truly uncon-
scious capacity that functions independent of conscious
perception. In the present study, we examine using signal
detection theoretic measures to what extent subliminal
perception is independent of subjective introspection.
Second, we investigate to what extent subliminal percep-
tion can be predicted based on the state of the brain
before stimulus presentation.
To a large degree, the debate whether subliminal
perception is a truly unconscious capacity stems from the
difficulty of determining subjective, conscious perception.
To measure whether a stimulus was consciously seen,
experimenters ask the participant to make a decision and
report what he or she saw. This means that the participant
must set some criterion level to which she compares the
strength of her subjective perception. When the “signal”
(i.e., conscious perception) falls below the criterion for
“not seen,” perception is considered unconscious. Subjec-
tive perceptual reports can be affected by multiple factors
(e.g., decision bias, perceptual bias, and motor bias;
Anzulewicz et al., 2019; Samaha et al., 2020; Siedlecka
et al., 2019). Different means of reporting conscious per-
ception can yield different results regarding the presence
of subliminal perception (Overgaard et al., 2006; Ramsøy &
Overgaard, 2004), indicating that at least sometimes sub-
liminal perception may only be severely degraded
conscious vision (Overgaard, 2011; Phillips, 2020).
According to the currently popular models of
vision, the feedforward sweep of stimulus-triggered
activation through cortical areas happens outside of
the individual’s consciousness, but this activation is
assumed to have the power to influence behavior
(Vanrullen, 2007; VanRullen & Koch, 2003). The extent
to which feedforward processes enable subliminal con-
tents to influence behavior may depend on selective
attention: attended stimuli may produce stronger activa-
tion, increasing the likelihood of correct behavioral
response (Dehaene et al., 2006). Feedforward activation
may also trigger the formation of recurrent activity loops
between different cortical areas, enabling stronger sub-
liminal perceptual capacity (Dehaene et al., 2006). These
conclusions are supported by studies showing that under
certain conditions, where the participants report not
seeing a stimulus, early visually evoked activation is
preserved but does not ignite widespread cortical
activation (Gaillard et al., 2009; Kouider et al., 2007;
Salti et al., 2015; Supèr et al., 2001). Widespread cortical
activation is assumed to enable full-fledged conscious
perception, allowing the participant to guide behavior
based on the conscious contents and, for example, report
what they perceived (i.e., “conscious access”; Dehaene &
Changeux, 2011; Lamme & Roelfsema, 2000).
If the accuracy of subliminal perceptual responses
depends on the strength of the early visual activation
(Dehaene et al., 2006), then visually evoked
activation should be higher in the trials where the
participants provide a correct answer. In line with this
prediction, some studies show enhanced event-related
potentials (ERP) around 200 ms after stimulus onset in
occipitotemporal electrodes when the participants
correctly discriminate a target they report not seeing
(Koivisto & Grassini, 2016; Lamy et al., 2009). However,
because reports of conscious vision correlate with ERP
amplitude in the same time window—an effect known as
visual awareness negativity (VAN; Förster et al., 2020;
Railo et al., 2011)—this correlate of subliminal percep-
tions could reflect degraded conscious perception rather
than strictly unconscious perception. Koivisto and
Grassini (2016) reported that the ERP correlate of sublim-
inal discrimination also correlated with conservative
response criterion, suggesting that what the participants
labeled “not seen” may have been severely degraded
conscious perception. This is consistent with the results
of Peters and Lau (2015), which suggest that when per-
ception is measured using a procedure that minimizes
the influence of criterion setting, no evidence for strictly
subliminal perception is observed (see also Rajananda
et al., 2020).
Preserved or enhanced early visually evoked neural
activation in subliminal trials may not alone explain why
in some subliminal trials the participant responds correctly
but not in other trials. In contrast to the idea that the
strength of early visually evoked responses correlates with
the likelihood of correct behavioral responses within
reportedly subliminal trials, Salti et al. (2012) found that
only the late P3 wave correlated with the accuracy of sub-
liminal discrimination (see also Salti et al., 2015). This
could mean that subliminal perception may be less
influenced by the strength of the visually evoked response
than currently assumed. Possibly the trial-by-trial variation
in behavioral performance on subliminal trials is more
related to how well decision-making processes can
“access” the visual stimulus-related information, rather
than how strong early activation the stimulus evokes.
Instead of focusing on visually evoked activity, we
reasoned that subliminal perception could be explained
by the state of neural activity before stimulus presenta-
tion. Neural activity preceding stimulus presentation may
influence how well the stimulus is perceived due to
multiple factors: The participants’ attentional state may
vary spontaneously, but it may also be under voluntary
control. Numerous studies have shown that conscious
4986 RAILO ET AL.
visual perception is strongly influenced by low-frequency
oscillations, especially the alpha (8–13 Hz) oscillations,
before stimulus presentation (Bareither et al., 2014;
Boncompte et al., 2016; Britz et al., 2014; Iemi &
Busch, 2018; Jensen et al., 2012; Kloosterman et al., 2019;
Limbach & Corballis, 2016; Mathewson et al., 2009;
Samaha et al., 2017; Schroeder & Lakatos, 2009; Thut
et al., 2006). However, prestimulus alpha power is cur-
rently assumed to influence conscious detection of stim-
uli and not perceptual discrimination or subliminal
perception (Benwell et al., 2017; Iemi et al., 2017; Samaha
et al., 2020).
In the current study, we randomly presented partici-
pants with a threshold-level low-contrast stimulus on
either the left or right visual field and asked them to
report in which visual hemifield the stimulus was pres-
ented and then rate how well they saw it (Figure 1a). We
reasoned that we would have the highest chance of
observing subliminal perception using this type of simple
location discrimination task. Moreover, we used
lateralized stimuli because we wanted to examine how
the pattern of lateralized prestimulus activation influ-
ences subliminal perception and evoked visual activity.
We used signal detection theoretic measures to test if
subliminal perception is independent of subjectively
reported visibility.
2 | MATERIALS AND METHODS
2.1 | Participants
Thirty-four students (3/34 males, 3/34 left-handed) with
a mean age of 24.4 (SD = 3.6) years and no neurological
disorders participated in the experiment. Three
participants were excluded before the EEG preprocessing
because of a high false alarm rate (participants with false
alarm rate >0.3 were excluded; mean false alarm rate of
the excluded participants was 0.59, SD = 0.26), and one
participant because she always reported seeing the stim-
uli. The participants were students at the University of
Turku. Each participant gave written informed consent.
The study was conducted in accordance with the Declara-
tion of Helsinki and was approved by the ethics commit-
tee of the Hospital District of Southwest Finland.
2.2 | Stimuli
The experiment was run in MATLAB (version R2014b)
using the Psychophysics Toolbox (Brainard, 1997). Stim-
uli were presented on an VIEWPixx/EEG LCD monitor
with a 120-Hz refresh rate. The stimuli were Gabor pat-
ches (a diameter of 6.5
 of visual angle and a frequency
of 0.7 cycles/degree; the phase and orientation of the
Gabor patch was randomly varied on each trial) pres-
ented on a gray 50% background (45.5 cd/m2) in left or
right hemifield (about 5
 from fixation on horizontal
meridian). The monitor was not gamma corrected, which
means that average luminance of the Gabor patch does
not precisely match the luminance of the background
(i.e., stimulus could be detected based on luminance or
contrast information). On two thirds of the trials, a low-
contrast Gabor patch was presented. The intensity of this
low-contrast Gabor was individually adjusted to be near a
50% subjective detection threshold using a QUEST stair-
case (Watson & Pelli, 1983). On one sixth of the trials, a
high-contrast Gabor with three times higher contrast
than the contrast of the low-contrast Gabor was pres-
ented. The high-contrast stimuli were presented to allow
the participant to clearly see the stimuli every now and
F I GURE 1 Behavioral paradigm and results. (a) A schematic of a single trial: the participants gave a forced discrimination response
and rated how well they saw the stimulus. (b) Correlation between the sensitivity of subjectively detecting the stimulus and proportion
correct (forced discrimination task) on subliminal trials. (c) Correlation between criterion for reporting conscious perception and
discrimination accuracy on subliminal trials
RAILO ET AL. 4987
then, meaning they could use the whole visibility rating
scale. Both low-contrast and high-contrast stimuli were
presented for a duration of two frames (16.6 ms in total).
One sixth of the trials did not have any stimulus (catch
trials), instead containing a blank screen with the fixation
point lasting two frames. Only trials with the low-
contrast stimuli were included in the statistical analysis.
2.3 | Experimental design
The stimulus was presented after a fixation period that
varied from 668 to 1332 ms from trial to trial to prevent
the participant from learning to expect the stimulus at a
specific delay; 250 ms after the Gabor was presented
(or catch trial), the fixation point turned into an arrow
pointing left and right, indicating that the participant
should try to report the side of the target. After this, the
numbers “0-1-2-3” were presented on the center of
the screen to prompt a visibility rating response. Both
responses were given by pressing a key on a numpad.
The participants were told to try to give their best guess
about stimulus location, even when they felt they did not
see any stimulus. The visibility rating (Overgaard, 2011;
Ramsøy & Overgaard, 2004; Sandberg et al., 2010) was
given using a four-step scale where the alternatives were
as follows: (0) “did not see stimulus at all,” (1) “not sure
but possibly saw something,” (2) “pretty sure I saw it,”
and (3) “saw the stimulus clearly.” The difference
between alternatives 0 and 1 was stressed to the partici-
pant. That is, we emphasized that if they felt that they
saw any glimpse of the stimulus, even an extremely weak
one, they should select the second lowest alternative. The
rating task was preferred over a yes/no task to make sure
that the participants used a strict criterion for reporting
no subjective perception of stimuli. Throughout the pre-
sent study, a participant is assumed to be subliminal of a
stimulus only when she/he chose the lowest visibility rat-
ing (the three higher alternatives are all counted as con-
scious). No instruction was given about the response
speed (of either response). The participants were also told
that in some trials, the stimulus would not be presented,
and they were told that on these trials, the correct
response was to report that they did not see the stimulus.
The experiment comprised a total of 400 trials per partici-
pant, subdivided into 10 blocks of 40 trials each.
2.4 | EEG recording and preprocessing
A 64-channel EEG was recorded at a 500-Hz sampling
rate with a NeurOne Tesla amplifier. Before the record-
ing, electrode impedances were brought near 5 kΩ. Two
additional electrodes were used to record electro-
oculograms (EOG): one electrode was placed beneath
and the other beside the participant’s left eye.
Preprocessing was carried out with MATLAB (ver-
sion R2016b) using the EEGlab toolbox (version 14.1.1b).
For each participant, bad channels (channels with no
response or very noisy signals) were manually checked
and interpolated. EEG data were re-referenced to the
average of all electrodes. A 0.25-Hz high-pass filter and
an 80-Hz low-pass filter were applied to the data. Line
noise was cleaned up using the CleanLine EEGlab plu-
gin. Data were epoched from 2000 to 1500 ms relative
to the stimulus onset. Trials with eye movements (EOG
electrodes) within 500 to 500 ms relative to the stimu-
lus onset were discarded (on average, 24 trials, SD = 37
trials were rejected per participant). The locations of the
electrodes were transposed so that the left hemisphere
was always contralateral and the right hemisphere
always ipsilateral relative to stimulus presentation
(we present nontransposed results in the Supporting
Information). A time frequency analysis was performed
on single trials using complex Morlet wavelets (fre-
quency range: 1–30 Hz with 2-Hz step size, with the
number of cycles increasing linearly from 1 to 12). No
baseline correction was applied when examining pre-
stimulus oscillations.
Due to the limited temporal resolution of the wavelet
analysis, poststimulus time windows could influence the
prestimulus data points, especially for low (<3 Hz) fre-
quencies. To control for this, we also calculated power
spectral densities from prestimulus (single trial) data seg-
ments. The EEG power spectrum reflects not only peri-
odic oscillatory signals but also aperiodic activity with
1/f-like distribution. We estimated exponent and offset of
aperiodic power across frequencies 1–40 Hz using fooof
package (Donoghue et al., 2020).
2.5 | Statistical analysis
Behavioral signal detection measures of sensitivity (d0)
and criterion (c) were calculated from binarized subjec-
tive visibility ratings. A hit was defined as a visibility rat-
ing of 1–3 when a stimulus was presented. False alarms
were trials where the participant reported seeing a target
(rating 1–3) when no stimulus was presented.
Influence of prestimulus EEG power on behavior was
examined as follows. First, time points between 1000 and
0 ms before stimulus onset in steps of 24 ms were statisti-
cally analyzed to examine if prestimulus oscillatory
power predicted behavioral performance. To see which
electrode clusters and frequencies may enable predicting
subliminal perception, these analyses were performed in
4988 RAILO ET AL.
a mass univariate manner (Maris & Oostenveld, 2007),
analyzing all the time points and all frequencies between
1 and 30 Hz. Statistical analyses were performed on the
single-trial data. We used mixed-effects logit models to
test whether log transformed oscillatory power at specific
time point and frequency predicted the accuracy of the
response (correct/incorrect) or subjective visibility (saw/-
did not see). The mixed-effects models included random
intercepts for participants. More complicated models
often failed to converge, and the models also took very
long to compute. Log transformed power was used as the
predictor because untransformed data were strongly right
skewed.
Statistical significance was assessed using permuta-
tion testing by randomly shuffling condition labels (1000
iterations). We used threshold-free cluster enhancement
(TFCE) to take into account the clustering of effects in
neighboring time points and frequency bands (Smith &
Nichols, 2009). Maps of the t values obtained from the
mass univariate analyses were TFCE transformed using a
function from the Limo EEG package (Pernet
et al., 2011), using default parameters (E = 0.5, H = 2,
dh = 0.1). TFCE was performed for the authentic data
and also after each permutation. The highest TFCE score
(TFCEmax) of each permutation was saved to obtain a
null distribution corrected for multiple comparisons. This
analysis was run separately on data from five clusters of
electrodes. To ensure two-tailed testing, and take into
account that analysis was repeated five times (for each
cluster of channels), we further used a Bonferroni
corrected alpha level to assess statistical significance.
TFCE values with the probability <0.005 when compared
with the TFCEmax distribution were considered statisti-
cally significant.
We also analyzed ERPs and event-related spectral
perturbations (ERSP) to examine how behavioral perfor-
mance (subjective and objective measures) was associated
with stimulus-evoked EEG responses. These analyses
were performed using a similar mass univariate approach
as described earlier (mixed-effects logit models with ran-
dom intercepts, 1000 permutations, and TFCE, and
p = 0.005 alpha level). In the ERP analyses, TFCE was
calculated over neighboring time samples and electrodes.
Before ERP analysis, EEG was low-pass filtered at 40 Hz.
Single-trial ERPs were baseline corrected to 200 ms. In
standard ERSP analysis, multiple trials are averaged, and
it is measured how EEG oscillatory power changes in
response to a stimulus relative to some baseline state. For
each trial, we subtracted from each poststimulus time
point and frequency the average prestimulus power
(at the corresponding frequency) between 1000 and
100 ms. To reduce computation time, the ERSP was res-
ampled to 24-ms steps for the statistical analysis.
The number of trials per participant in the statistical
analyses is as follows. Participants had on average 125 tri-
als in the conscious condition (SD = 42 trials), and
120 (SD = 39) trials in the subliminal condition. When
both seen and unseen trial are included in the analysis,
the comparison of correct versus incorrect objective dis-
crimination is based on 190 (SD = 37) versus
55 (SD = 24) trials per participant on average, respec-
tively. The main analysis of objective discrimination per-
formance on subliminal trials is based on average on
71 (correct response, SD = 23) versus 49 (incorrect
response, SD = 20) trials per participant.
Datasets are available for download at https://osf.io/
xz8jr/.
3 | RESULTS
3.1 | Behavioral results
Three participants displayed a very strong rightward
response bias: when they reported not seeing a stimulus,
they always pressed that the stimulus was presented on
the right side. Because the aim of this study was to see if
participants who report not seeing a stimulus correctly
guess the location of the stimulus, these participants were
excluded from the behavioral and EEG analysis. The
remaining participants did not, as a group, show system-
atic lateral biases: when no stimulus was presented, these
participants responded that the stimulus was on the right
hemifield on average on 52% of trials (SD = 2%; one sam-
ple t test against 50% level: t = 0.74, df = 26, p = 0.46).
Visibility ratings of left and right hemifield targets did
not differ statistically significantly (t = 1.13, df = 26,
p = 0.26). That these analyses are based on aggregated
data, not single trials as the EEG results presented below.
On average, the participants reported seeing the tar-
get (visibility rating 1–3) in 57% of the trials (SEM = 3%)
when the stimulus was presented. When the stimulus
was not presented, they rarely reported seeing the target
(mean = 19%, SEM = 4%). When the participants
reported not seeing the target at all (visibility rating = 0),
they nevertheless correctly reported the side of the stimu-
lus in 59% of the trials on average (t test against 50%
chance-level: t = 6.14, df = 26, p = 1.71  106)—we
refer to this as “subliminal perception.”
Do accurate responses in the trials where the partici-
pants report no conscious perception reveal the existence
of a perceptual capacity that functions independently of
conscious vision? We calculated the participants’ sensitiv-
ity and criterion for consciously detecting the target
(i.e., the measures were calculated from subjective
reports, not objective performance, using signal detection
RAILO ET AL. 4989
analyses). As shown in Figure 1b, the participants who
were not sensitive to subjectively detecting the stimulus
(d0 ≈ 0) did not reveal subliminal perception (intercept of
the linear model = 0.49; effect of d0, β = 0.072, t = 5.17).
As shown in Figure 1c, the participants also tended to
use a conservative criterion for reporting conscious per-
ception. This is common with low-contrast stimuli
because the participants try to minimize false alarms.
Based on extrapolating the fitted linear model, if the par-
ticipants used statistically optimal criterion (c = 0) for
reporting consciousness, they would also have performed
at near the chance level in discriminating the stimulus
side (intercept = 0.54; effect of criterion, β = 0.098,
t = 3.76). Altogether, these results suggest that sublimi-
nal perception was not separate from subjectively
reported vision.
3.2 | Prestimulus EEG correlates
We tested if prestimulus EEG power predicted the partic-
ipants’ objective discrimination performance when they
subjectively reported not seeing the target at all (the low-
est visibility rating). Because the lateralization of cortical
excitability varies depending on whether attention is
focused toward the right or left visual hemifield
(Boncompte et al., 2016; Thut et al., 2006; Worden
et al., 2000), we took this into account by transposing the
electrode locations for each trial so that the left electrode
locations represent electrodes that were contralateral
with respect to the visual stimulus (and right electrodes
were ipsilateral). The electrode clusters are visualized in
Figure 2a.
As shown in Figure 2b, objective performance on trials
where the participants reported not seeing the stimulus at
all could be predicted from prestimulus power. Oscillatory
power at the occipital electrode clusters predicted the
accuracy of responses almost up to 1 s before the stimulus
was presented, even though the participants reported just
guessing. The effect is lateralized: oscillatory power 8–
12 Hz contralateral to the stimulus and 1–5 Hz in ipsilat-
eral occipital electrodes predicts better performance. Sta-
tistically significant effects are observed in the occipital
electrode clusters, but the frontal and central electrode
clusters showed a similar pattern of effects. When elec-
trode locations were not transposed, a weaker prestimulus
effect in the alpha band in occipital and central electrodes
(up to 250 ms before stimulus onset) on unconscious dis-
crimination was observed (Figure S1).
The prestimulus power in single trials was also associ-
ated with reported conscious visibility of the target in a
trial-by-trial manner (Figure 2c), but the effect was clearly
smaller when compared with effects on unconscious
discrimination. A similar effect is found when non-
transposed electrode locations were used (Figure S2).
To rule out the possibility that the observed pre-
stimulus correlates (Figure 2b) were caused by temporal
smearing of the wavelet, we calculated power spectral den-
sity from single-trial prestimulus data segments (500 to
24 ms). We then ran a mixed-effects logit model where
we predicted subliminal discrimination accuracy in single
trials based on the average power spectral density of the
ipsilateral electrode cluster (1–3 Hz) and the average
spectral density in the contralateral electrode cluster
(8–12 Hz). Oscillatory power in contralateral occipital
electrode clusters predicted subliminal discrimination
performance, replicating previous analyses (t = 3.05).
Ipsilateral occipital power did not predict subliminal dis-
crimination accuracy (t = 1.30). The association between
contralateral/ipsilateral occipital power spectral density
and subliminal perception are visualized in Figure 3a,b.
Next, we examined how well oscillatory power at
these two frequencies/electrode clusters predict sublimi-
nal discrimination accuracy when confounding variables
are controlled for. First, we included in the model the
hemifield in which the stimulus was presented to control
for the possibility that the effect was caused by a spatial
bias. Interaction between stimulus hemifield and contra-
lateral or ipsilateral power are not included in the model
presented below because these did not improve the
model. Second, accuracy of previous trial’s (whether sub-
liminal or conscious) response was included in the model
to account for temporal correlations throughout the
experiment (see, e.g., Schaworonkow et al., 2015). Predic-
tors related to trial number, participant’s response in the
previous trial (left/right), and the stimulus hemifield in
the previous trial were initially included in the model but
left out because they failed to explain variation in sublim-
inal discrimination. Finally, to this model, we added
stimulus hemifield, contralateral power, and ipsilateral
power as participant-wise random-effect predictors to
take into account individual variation in these factors. As
shown in Table 1, subliminal discrimination was more
accurate in the right hemifield (t = 2.06). The participant
was more likely to correctly discriminate the subliminal
stimulus when he/she responded correctly in the previ-
ous trial (t = 3.89). Even when these factors were
included in the model, contralateral occipital prestimulus
power predicted subliminal discrimination accuracy
(t = 2.35). Ipsilateral power did not statistically signifi-
cantly predict subliminal discrimination accuracy in this
model (t = 1.21). The area-under-the-curve (AUC) of
this model was 0.70. The same model with shuffled labels
produced a maximum AUC of 0.60 (out of 1000 permuta-
tions), and a model with only the intercept term had an
AUC of 0.58.
4990 RAILO ET AL.
The above results suggest that behavioral perfor-
mance in subliminal trials correlates with prestimulus
power in specific low-frequency bands. To examine
whether these effects stem from true oscillatory activity,
or changes in broadband aperiodic activity (Donoghue
et al., 2020), we examined to what extent markers of ape-
riodic EEG activity predict subliminal discrimination
accuracy. The aperiodic components (exponent and
offset) were calculated separately for occipital ipsilateral
and contralateral electrode clusters (from prestimulus
data segments). Table 2 shows to what extent the markers
of aperiodic activity predicted subliminal discrimination.
Neither contralateral alpha (t = 1.34) or ipsilateral delta
(t = 0.30) oscillatory power predicted subliminal dis-
crimination accuracy when they were included in the
model together with the other predictors. In addition to
F I GURE 2 How prestimulus oscillatory power predicts behavior. (a) Electrode clusters used in EEG analysis. Each color represents
one cluster. Plus-symbols are electrodes that are not part of any cluster. (b) Results of mass univariate mixed-effects logit models where a
single trial’s prestimulus EEG power was used to predict the location discrimination accuracy on subliminal trials (visibility = 0) and
(c) binarized visibility ratings (did not see = 0; 1–3 = saw). Color is the t value of the mixed-effects logit model. The blue color denotes a
negative correlation and red a positive correlation. The faint colors are time points/frequencies where the model does not statistically
significantly predict behavior. Bright colors show statistically significant modulation according to TFCE and permutation testing. Frontal
channel clusters are not shown because no statistically significant effects were observed
F I GURE 3 Influence of contralateral and ipsilateral occipital power on subliminal discrimination performance. Panels (a) shows the
contralateral electrode cluster, and panel (b) the ipsilateral electrode cluster. Line and strong data points show the average across
participants, and faint symbols individual participants’ data (average across trials). Panels (c) and (d) show the average power spectrum in
contralateral and ipsilateral electrode clusters, respectively. Black line is the condition where the location of a subliminal target was correctly
discriminated. Red is incorrect discrimination on subliminal trials
RAILO ET AL. 4991
this “full model” (BIC = 13,652), a pruned model
(BIC = 13,636) that does not include these two predictors
is presented in Table 2. In the pruned model, only the off-
set of estimated aperiodic spectrum predicted subliminal
discrimination: weaker contralateral (t = 2.11) and
stronger ipsilateral (t = 1.87) broadband activity
predicted subliminal discrimination. This finding is
important because it suggests that the prestimulus EEG
feature that predicts subliminal discrimination may not
be an oscillation but the strength of nonoscillatory
broadband EEG.
We also examined if occipital prestimulus power spec-
tral density data predicted reported stimulus visibility.
First, we tested if contralateral power and ipsilateral
power (same frequencies and time windows as previ-
ously) predicted stimulus visibility when confounding
variables are included in the model. The results of the
model (which included stimulus hemifield, contralateral
power, and ipsilateral power as participant-wise random-
effect predictors) are presented in Table 3. Here, ipsilat-
eral prestimulus power was negatively associated with
reported stimulus visibility (t = 2.25), and again,
correct discrimination of the target in previous trial
predicted higher likelihood of conscious detection
(t = 3.08). Contralateral power did not predict conscious
perception (t = 0.49). This is noteworthy because in the
comparable model concerning subliminal discrimination
(Table 1), contralateral (but not ipsilateral) occipital
prestimulus power predicted better performance. This
suggests that different prestimulus neural processes may
mediate subliminal discrimination and subjective visibil-
ity (although based on our behavioral results, they both
lie on the same continuum of conscious vision).
Aperiodic prestimulus activity did not predict
stimulus visibility, and hence, the results are not pres-
ented further (in a model with four aperiodic predictors,
highest absolute t value = 0.6).
3.3 | Event-related potentials
What is the mechanism through which prestimulus
power influences the discrimination of subliminal
stimuli? High-amplitude low-frequency oscillations have
TAB L E 1 Results of a mixed-effects logit model predicting subliminal discrimination accuracy (df = 3291)
Name Estimate SE t p CI (lower) CI (upper)
Intercept 0.322 0.214 1.500 0.134 0.742 0.099
Contralateral power 0.043 0.018 2.359 0.018 0.078 0.007
Ipsilateral power 0.021 0.017 1.217 0.224 0.013 0.054
Stimulus in right hemifield 0.833 0.404 2.060 0.039 0.040 1.626
Accuracy in previous trial 0.342 0.088 3.895 <0.001 0.170 0.515
TAB L E 2 Results of mixed-effects logit models predicting subliminal discrimination accuracy (df = 3182)
Name Estimate SE t p CI (lower) CI (upper)
Full model
Intercept 0.357 0.221 1.613 0.107 0.077 0.790
Contralateral exponent 0.012 0.067 0.172 0.863 0.120 0.143
Ipsilateral exponent 0.053 0.064 0.839 0.402 0.178 0.071
Contralateral offset 0.068 0.059 1.162 0.245 0.184 0.047
Ipsilateral offset 0.110 0.060 1.832 0.067 0.008 0.228
Contralateral power 0.025 0.019 1.341 0.180 0.062 0.012
Ipsilateral power 0.006 0.018 0.307 0.759 0.041 0.030
Pruned model
Intercept 0.340 0.224 1.518 0.129 0.099 0.779
Contralateral exponent 0.051 0.060 0.849 0.396 0.067 0.169
Ipsilateral exponent 0.041 0.059 0.693 0.488 0.155 0.074
Contralateral offset 0.108 0.051 2.118 0.034 0.209 0.008
Ipsilateral offset 0.093 0.050 1.877 0.061 0.004 0.191
4992 RAILO ET AL.
been associated with suppressing the processing of nonre-
levant signals, whereas low-amplitude low-frequency
power is associated with amplifying sensory input
(Schroeder & Lakatos, 2009). The lateralized contribution
of occipital prestimulus power on subliminal discrimina-
tion performance suggests that successful behavioral
responding in subliminal trials may be related to
enabling strong stimulus-evoked responses to the (contra-
lateral) visual stimulus while suppressing unwanted sig-
nals in the opposite (ipsilateral) visual field (where no
target is presented). If this is the case, we should observe
enhanced visually evoked responses to correctly discrimi-
nated subliminal targets (when compared with the trials
of incorrect responses).
To examine how subliminal discrimination accuracy
and subjective visibility of the stimulus influences
stimulus-evoked activity, we analyzed ERPs in a single-
trial manner. As in the prestimulus analyses, the electrode
locations in these analyses were transposed so that in the
scalp maps, the left hemisphere locations correspond to
the contralateral hemifield. As shown in Figure 4a, subjec-
tive visibility of the stimulus was associated with a strong
increase of ERP amplitudes 275 ms after stimulus onset,
spanning most electrode locations. Nearly identical effects
are produced by when objective discrimination perfor-
mance is examined (all trials included; Figure S3), in line
with the behavioral result that objective discrimination
was not independent of subjectively reported vision. How-
ever, as shown in Figure 4b, no strong, statistically signifi-
cant effects are seen in ERPs when the analysis is
restricted to subliminal trials. Note however that Figure 4b
suggests that ERP amplitudes are more negative in contra-
lateral occipital electrodes around 175–275 ms, although
this effect does not reach statistical significance in mass
univariate testing.
Similar pattern was revealed by event-related spectral
perturbations (ERSP): even though stimulus visibility
clearly modulated ERSPs (Figure S4), discrimination
accuracy on subliminal trials did not statistically signifi-
cantly modulate ERSPs at occipital electrodes
(Figure S5).
Next, we examined how prestimulus power
influenced ERPs. If the prestimulus process that mediates
accurate subliminal discrimination relates to fluctuations
in cortical excitability, we should be able to observe that
the contralateral occipital power is inversely related to
ERP. Because the prestimulus EEG analyses indicated
that broadband prestimulus offset correlated with sub-
liminal discrimination, we first examined if this predicted
ERPs (mixed-effects logit models with two predictors
corresponding to offset in ipsilateral and contralateral
occipital channels). These analyses did not yield any sta-
tistically significant effects.
We also examined whether single-trial oscillatory
power in the same frequencies/time windows that
predicted objective behavioral performance (ipsilateral
electrode cluster at 1–3 Hz and contralateral electrode
cluster at 8–12 Hz, 500–24 ms before stimulus onset)
predicted ERP amplitude. We included only subliminal
trials in the analysis. The results are shown in Figure 5.
Contralateral occipital prestimulus power predicted ERP
amplitude around 275 ms after stimulus onset in frontal
electrodes. This effect is consistent with the observed
relationship between ERPs and behavioral performance:
As shown in Figure 4, conscious perception is associated
with more positive ERPs in frontal sites after 275 ms.
Figure 5 shows that the higher contralateral occipital pre-
stimulus power, the weaker the ERP is in frontal elec-
trodes around 275 ms. Contralateral prestimulus power
also statistically significantly modulated later responses
(500 ms) in contralateral frontal electrodes. Ipsilateral
occipital prestimulus power did not predict ERP
amplitudes.
4 | DISCUSSION
We investigated what factors explain why behavioral
visual discrimination accuracy can be on average above
the level of chance, even though the participants subjec-
tively report not seeing the stimulus (i.e., “subliminal per-
ception”). Our results reveal the following: first,
subliminal perceptual performance was not completely
independent of conscious introspection, as measured by
the signal detection theory. Second, lateralized neural
activity before stimulus presentation, lateralization of
TAB L E 3 Results of a mixed-effects logit model predicting stimulus visibility (df = 6846)
Name Estimate SE t p CI (lower) CI (upper)
Intercept 0.204 0.132 1.548 0.122 0.462 0.054
Contralateral power 0.008 0.016 0.493 0.622 0.039 0.023
Ipsilateral power 0.030 0.013 2.250 0.024 0.055 0.004
Accuracy in previous trial 0.192 0.062 3.080 0.002 0.070 0.315
Stimulus in right hemifield 0.032 0.081 0.392 0.695 0.127 0.191
RAILO ET AL. 4993
stimuli, and previous trial’s discrimination accuracy
predicted behavioral performance even when the partici-
pants reported just making guesses. Third, the accuracy
of subliminal discrimination did not clearly modulate
stimulus-evoked responses (ERP or ERPS), although as
discussed below, this result may indicate lack of statisti-
cal power.
4.1 | Subliminal discrimination was not
based on completely unconscious signals
Our behavioral results suggest that the performance on
reportedly subliminal trials was not completely indepen-
dent of the participants’ introspective abilities, at least
based on the signal detection theoretic metrics of
F I GURE 4 Event-related potentials. (a) Single-trial visibility ratings (top panels) predicted ERPs 275 ms after stimulus onset. (b) When
only reportedly subliminal trials were included in the analysis (bottom panels), no strong effects were observed. The figures on the left show
the results of the mass univariate mixed-effects regression models. Color denotes t value, and statistically significant effects (based on TFCE)
are shown with a bright color. The middle panels show scalp topographies of the effects at 276 and 400 ms after stimulus onset (note that the
electrode locations have been transposed: electrodes over the left scalp locations are contralateral to the stimulus). The rightmost panels
show ERPs measured at the contralateral occipital cluster. The gray line depicts consciously seen (top panels) or correctly discriminated
(bottom panels) trials, and the black line depicts not consciously seen (top panels)/incorrectly discriminated trials (bottom panels)
F I GURE 5 Influence of single-trial
prestimulus oscillatory power on stimulus-
evoked ERPs. Effect of contralateral occipital
alpha prestimulus power on ERPs when the
analysis includes only subliminal trials. The
rectangular color map plot shows the results of
the mass univariate linear mixed-effects
models. Color represents the t value. A strong
color represents statistically significant effects,
as determined by TFCE. Scalp maps show the
distribution of the effects at specified time
points (note that the electrode locations have
been transposed: electrodes over the left
hemisphere are contralateral to the stimulus)
4994 RAILO ET AL.
sensitivity and criterion (which were calculated based on
subjective perceptual reports). This means that when the
participants reported not seeing the stimuli, the above
chance behavioral performance on these subliminal trials
was likely nevertheless based on signals that the partici-
pants could introspect and base their response to. While
it could be debated whether the nature of the information
on what the participants based their decision on was
visual or not (perhaps the experiences were type-2 like,
as sometimes reported by blindsight patients), the key
point is that the participants had some type of conscious
access to information about the location of the stimulus
on subliminal trials.
In general, signal detection theoretic framework
implies that dissociations between subjective detection
and forced-choice discrimination may be observed, not
because there is a distinct unconscious perceptual capac-
ity but because the sensitivity of discrimination task is
higher than the sensitivity of perceptual reports
(Campion et al., 1983; King & Dehaene, 2014;
Phillips, 2018). Part of the difficulty in answering
whether the signals on which the participants based their
responses on subliminal trials were unconscious or not is
that it requires us to artificially binarize a process that
is in truth continuous. Our finding is in line with the
observation that when a paradigm is used which mini-
mizes the participants’ need to specify a criterion to label
something as “conscious” or “unconscious,” no evidence
for strictly unconscious perception is observed (Peters &
Lau, 2015; Rajananda et al., 2020). Also, objective visual
discrimination behavior and subjectively experienced
perception evoked nearly identical ERP/ERSP responses,
suggesting that the two are not strongly dissociated.
Our results are interesting when considered from the
perspective of blindsight patients. Blindsight is typically
assumed to demonstrate the existence of anatomical
pathways that relay unconscious information to cortex,
enabling the patients to use stimuli presented to their
blind visual field to guide behavior (Ajina et al., 2015;
Ajina & Bridge, 2018; Azzopardi & Cowey, 1997;
Cowey, 2010; Danckert & Rossetti, 2005; Weiskrantz
et al., 1974). Whether blindsight patients’ behavioral
responding is based on completely unconscious percepts
(Campion et al., 1983; Phillips, 2018), and whether
similar mechanisms guide behavior in neurologically
healthy individuals remains debated (Allen et al., 2020;
Hurme et al., 2017, 2019, 2020; Lloyd et al., 2013; for a
review, see Railo & Hurme, 2021). An interesting open
question is to what extent prestimulus processes also
contribute to blindsight and explain trial-by-trial
variation in blindsight performance.
Finally, our results indicate that subliminal percep-
tion was more likely to be correct if the participant
correctly discriminated the target in the previous trial.
Furthermore, correct subliminal discrimination was
more likely when stimuli were presented in the right
hemifield. To what extent these factors reveal response
biases, attentional or decision-making biases, or long-
term autocorrelations in the data (Schaworonkow
et al., 2015) is unclear. Altogether, our results indicate
that many different factors contribute to discrimination
accuracy on reportedly subliminal trials. Subliminal dis-
crimination accuracy could be predicted with
AUC = 70% accuracy in the present study, which is very
good performance given that we are essentially classify-
ing trials which participants themselves label as
“guesses.”
4.2 | Electrophysiological correlates
Baria et al. (2017) showed that participants’ behavioral
responses on subliminal trials could be predicted from
prestimulus activity patterns: when prestimulus activity
resembled left stimulus-evoked responses, the partici-
pants were more likely to respond “left.” Because we
transposed our electrode locations to correspond to con-
tralateral versus ipsilateral space, our analysis was sensi-
tive to the spatial pattern of the evoked response
(although this also means that we may miss effects that
are lateralized predominantly to left/right hemisphere).
Our motivation to transpose electrode locations to contra-
lateral versus ipsilateral space was largely that the lateral-
ization of prestimulus power could explain how strongly
the visual system responds to a visual stimulus in the
contralateral visual field due to variation in cortical excit-
ability. As discussed below, to what extent the present
results are consistent with the “cortical excitability”
explanation can be debated.
While previous research suggests that prestimulus
alpha power modulates subjective detection and not
objective discrimination (Iemi et al., 2017; Samaha
et al., 2020), our results show that when the task is to dis-
criminate lateralized targets, prestimulus alpha may also
influence objective discrimination performance. In our
study, we did not observe strong prestimulus effects for
subjective visibility. However, as seen in Figures 2 and
S2, subjective visibility was inversely associated with pre-
stimulus alpha, although this effect did not quite reach
statistical significance. Our results show that it is impor-
tant to take into account the lateralization of prestimulus
activity when examining prestimulus influences on per-
ception when the task involves discrimination across
hemifields. Although we observe that prestimulus activity
predicts discrimination performance, the present results
do not have to imply that prestimulus activity modulated
RAILO ET AL. 4995
perceptual sensitivity per se. Rather, because we trans-
posed electrode locations, our analysis is sensitive to
baseline excitability shifts (see Iemi et al., 2017; Samaha
et al., 2020) in both the contralateral and the ipsilateral
hemispheres. While our results consistently indicate that
contralateral occipital prestimulus activity was an impor-
tant predictor of subliminal discrimination, also ipsilat-
eral occipital activity may contribute to subliminal
perception. Note that control analyses indicated that the
prestimulus effects observed in the present study were
possibly mediated by aperiodic broadband activity
instead of oscillatory activity in the alpha band.
Early ERP waves (<200 ms after stimulus onset) are
assumed to reflect the addition of sensory-evoked neural
activation on top of ongoing neural activity (Mäkinen
et al., 2005; Shah et al., 2004). In addition to this, later
ERP responses are modulated by asymmetric amplitude
modulation of brain oscillations (Mazaheri &
Jensen, 2008; Nikulin et al., 2007). Prestimulus neural
activity may modulate ERPs through both of these mech-
anisms. Iemi et al. (2019) showed that while the power of
prestimulus alpha- and beta-band activity influenced the
earliest visual ERPs (C1 and N1 waves), the asymmetry
of oscillations predominantly influenced later ERP
waves.
We did not observe effects on early ERP amplitudes
in the present study. One possibility is that our result is
true, and there were not strong differences between early
evoked activity in the correct versus incorrect subliminal
discrimination trials. This may suggest that correct
responding in subliminal trials is determined by higher
level decision-related processes rather than (relatively
early) sensory processes. Interestingly, contralateral
occipital prestimulus power predicted ERP amplitude in
frontal areas when only the subliminal trials were
included in the analysis. The topography and timing of
this effect resembles correlates of conscious vision, and
P3a ERP component, which has been interpreted as a
correlate of orienting attention toward stimuli
(Polich, 2007). This suggests that correct discrimination
on subliminal trials may have to do with successfully
attending the conscious “first-order” stimulus-related
representations to enable correct discrimination decision
(see Ko & Lau, 2012, for related mechanisms in
blindsight patients). Prefrontal decision-making mecha-
nisms can be activated by subliminal stimuli (Lau &
Passingham, 2007; Van Gaal et al., 2010). Studies have
also reported that activity in the frontal cortical areas cor-
relates with subliminal behavioral performance (Lau &
Passingham, 2006) and conscious vision (Persaud
et al., 2011) in blindsight patients.
The influence of prestimulus oscillatory power on
ERP amplitude could be a confound resulting from
baseline correction of ERP’s: trials with high prestimulus
alpha power will also have nonzero mean amplitude due
to the asymmetrical amplitude of the oscillations (Iemi
et al., 2019). Therefore, it could be argued that when
amplitude asymmetry is high (high alpha power), high
amplitudes will be subtracted from the single-trial ERPs
during baseline correction. Note that similar confound
could also influence the ERP correlates of conscious and
subliminal perception, either masking true effects or pro-
ducing spurious findings.
It also remains possible that the strength of stimulus-
evoked activity was modulated in the present study, but
we were not able to detect this (likely very small) change
due to, for example, limited statistical power of the mass
univariate testing. In fact, Figure 4b suggests that correct
subliminal perception was associated with amplified neg-
ative ERPs in contralateral occipital electrodes around
200 ms after stimulus onset. As noted, conscious percep-
tion of stimulus often correlates with ERPs in the same
time window (termed “VAN”; Förster et al., 2020),
suggesting that correct subliminal discrimination is based
on degraded conscious vision (Koivisto & Grassini, 2016).
Similarly, while we did not observe statistically signifi-
cant VAN in the present research, this may be due to lack
of statistical power (due to mass-univariate testing) than
lack of real amplitude difference in VAN time window.
Because the aim of the present research was on sub-
liminal perception, we restricted analysis to trials where
participants reported not seeing the stimulus. An alterna-
tive method would have been to look for an interaction
between, for example, stimulus visibility and EEG mea-
sures (e.g., Benwell et al., 2017; Samaha et al., 2017). A
demonstration of an interaction that reportedly visible
and invisible trials are associated with different electro-
physiological correlates could indicate that conscious and
subliminal perception rely on clearly separate neural
mechanisms. However, it is also possible that, for exam-
ple, the neural correlates of above-threshold perception
and near-threshold perception differ, not because these
are truly different forms of perception but simply because
they are two different perceptual decisions based on dif-
ferent perceptual contents (e.g., one based on clear per-
ception of a stimulus and the other based on severely
degraded percept).
4.3 | Conclusion
We have shown that neural activity before stimulus pre-
sentation can be used to predict whether an individual
will correctly discriminate the location of a stimulus they
report not seeing. Furthermore, our results indicate that
performance on these reportedly unconscious trials may
4996 RAILO ET AL.
actually depend on processes that a participant can con-
sciously introspect. This way, our finding speaks against
the popular notion of a separate, unconscious perceptual
capacity, and emphasizes that visual behavior forms a
continuum that is tightly linked with the associated sub-
jective feeling of seeing.
ACKNOWLEDGMENTS
The study was funded by the Academy of Finland (grant
#308533). The authors also thank two reviewers for very
helpful comments.
CONFLICT OF INTEREST
No conflicts of interest.
AUTHOR CONTRIBUTIONS
H.R. designed the experiment, supervised data collection,
and wrote the manuscript. R.P. and H.R. analyzed the
data. R.P. and K.L. collected the data and commented
the manuscript.
PEER REVIEW
The peer review history for this article is available at
https://publons.com/publon/10.1111/ejn.15354.
DATA AVAILABILITY STATEMENT
Datasets are available at https://osf.io/xz8jr/.
ORCID
Henry Railo https://orcid.org/0000-0002-0779-5008
REFERENCES
Ajina, S., & Bridge, H. (2018). Blindsight relies on a functional
connection between hMT+ and the lateral geniculate nucleus,
not the pulvinar. PLoS Biology, 16, e2005769. https://doi.org/
10.1371/journal.pbio.2005769
Ajina, S., Pestilli, F., Rokem, A., Kennard, C., & Bridge, H. (2015).
Human blindsight is mediated by an intact geniculo-
extrastriate pathway. eLife, 4, e08935. https://doi.org/10.7554/
eLife.08935
Allen, C., Viola, T., Irvine, E., Sedgmond, J., Castle, H., Gray, R., &
Chambers, C. D. (2020). Causal manipulation of feed-forward
and recurrent processing differentially affects measures of
consciousness. Neuroscience of Consciousness., 2020, niaa015.
https://doi.org/10.1093/nc/niaa015
Anzulewicz, A., Hobot, J., Siedlecka, M., & Wierzchon, M. (2019).
Bringing action into the picture. How action influences visual
awareness. Attention, Perception, & Psychophysics, 81(7),
2171–2176. https://doi.org/10.3758/s13414-019-01781-w
Azzopardi, P., & Cowey, A. (1997). Is blindsight like normal,
near-threshold vision? Proceedings of the National Academy of
Sciences of the United States of America, 94, 14190–14194.
https://doi.org/10.1073/pnas.94.25.14190
Bareither, I., Chaumon, M., Bernasconi, F., Villringer, A., &
Busch, N. A. (2014). Invisible visual stimuli elicit increases
in alpha-band power. Journal of Neurophysiology, 112,
1082–1090. https://doi.org/10.1152/jn.00550.2013
Baria, A. T., Maniscalco, B., & He, B. J. (2017). Initial-state-
dependent, robust, transient neural dynamics encode con-
scious visual perception. PLOS Computational Biology, 13(11),
e1005806. https://doi.org/10.1371/journal.pcbi.1005806
Benwell, C. S. Y., Tagliabue, C. F., Veniero, D., Cecere, R.,
Savazzi, S., & Thut, G. (2017). Prestimulus EEG power predicts
conscious awareness but not objective visual performance.
ENeuro, 4, ENEURO.0182-17.2017. https://doi.org/10.1523/
ENEURO.0182-17.2017
Boncompte, G., Villena-Gonzalez, M., Cosmelli, D., & Lopez, V.
(2016). Spontaneous alpha power lateralization predicts
detection performance in an un-cued signal detection task.
PLoS ONE, 11, e0160347. https://doi.org/10.1371/journal.pone.
0160347
Brainard, D. H. (1997). The psychophysics toolbox. Spatial Vision,
10, 433–436. https://doi.org/10.1163/156856897X00357
Britz, J., Hernàndez, L. D., Ro, T., & Michel, C. M. (2014). EEG-
microstate dependent emergence of perceptual awareness.
Frontiers in Behavioral Neuroscience, 8, 163. https://doi.org/10.
3389/fnbeh.2014.00163
Campion, J., Latto, R., & Smith, Y. M. (1983). Is blindsight an effect
of scattered light, spared cortex, and near-threshold vision?
Behavioral and Brain Sciences, 6, 423–448. https://doi.org/10.
1017/S0140525X00016861
Cowey, A. (2010). The blindsight saga. Experimental Brain Research,
200, 3–24. https://doi.org/10.1007/s00221-009-1914-2
Danckert, J., & Rossetti, Y. (2005). Blindsight in action: What can
the different sub-types of blindsight tell us about the control of
visually guided actions? Neuroscience and Biobehavioral
Reviews, 29, 1035–1046. https://doi.org/10.1016/j.neubiorev.
2005.02.001
Dehaene, S., & Changeux, J. P. (2011). Experimental and theoretical
approaches to conscious processing. Neuron, 70, 200–227.
https://doi.org/10.1016/j.neuron.2011.03.018
Dehaene, S., Changeux, J. P., Naccache, L., Sackur, J., & Sergent, C.
(2006). Conscious, preconscious, and subliminal processing: A
testable taxonomy. Trends in Cognitive Sciences, 10, 204–211.
https://doi.org/10.1016/j.tics.2006.03.007
Donoghue, T., Haller, M., Peterson, E. J., Varma, P., Sebastian, P.,
Gao, R., Noto, T., Lara, A. H., Wallis, J. D., Knight, R. T.,
Shestyuk, A., & Voytek, B. (2020). Parameterizing neural
power spectra into periodic and aperiodic components. Nature
Neuroscience, 23(12), 1655–1665. https://doi.org/10.1038/
s41593-020-00744-x
Förster, J., Koivisto, M., & Revonsuo, A. (2020). ERP and MEG
correlates of visual consciousness: The second decade.
Consciousness and Cognition, 80, 102917. https://doi.org/10.
1016/j.concog.2020.102917
Gaillard, R., Dehaene, S., Adam, C., Clémenceau, S., Hasboun, D.,
Baulac, M., Cohen, L., & Naccache, L. (2009). Converging
intracranial markers of conscious access. PLoS Biology, 7, e61.
https://doi.org/10.1371/journal.pbio.1000061
Hannula, D. E., Simons, D. J., & Cohen, N. J. (2005). Imaging
implicit perception: Promise and pitfalls. Nature Reviews
Neuroscience, 6, 247–255. https://doi.org/10.1038/nrn1630
Hurme, M., Koivisto, M., Henriksson, L., & Railo, H. (2020). Neu-
ronavigated TMS of early visual cortex eliminates unconscious
RAILO ET AL. 4997
processing of chromatic stimuli. Neuropsychologia, 136, 107266.
https://doi.org/10.1016/j.neuropsychologia.2019.107266
Hurme, M., Koivisto, M., Revonsuo, A., & Railo, H. (2017). Early
processing in primary visual cortex is necessary for conscious
and unconscious vision while late processing is necessary only
for conscious vision in neurologically healthy humans.
NeuroImage, 150, 230–238. https://doi.org/10.1016/j.
neuroimage.2017.02.060
Hurme, M., Koivisto, M., Revonsuo, A., & Railo, H. (2019). V1
activity during feedforward and early feedback processing is
necessary for both conscious and unconscious motion percep-
tion. NeuroImage, 185, 313–321. https://doi.org/10.1016/j.
neuroimage.2018.10.058
Iemi, L., & Busch, N. A. (2018). Moment-to-moment fluctuations in
neuronal excitability bias subjective perception rather than
strategic decision-making. ENeuro., 5, ENEURO.0430-17.2018.
https://doi.org/10.1523/ENEURO.0430-17.2018
Iemi, L., Busch, N. A., Laudini, A., Haegens, S., Samaha, J.,
Villringer, A., & Nikulin, V. V. (2019). Multiple mechanisms
link prestimulus neural oscillations to sensory responses. eLife,
8, e43620. https://doi.org/10.7554/eLife.43620
Iemi, L., Chaumon, M., Crouzet, S. M., & Busch, N. A. (2017). Spon-
taneous neural oscillations bias perception by modulating
baseline excitability. Journal of Neuroscience, 37, 807–819.
https://doi.org/10.1523/JNEUROSCI.1432-16.2016
Jensen, O., Bonnefond, M., & VanRullen, R. (2012). An oscillatory
mechanism for prioritizing salient unattended stimuli. Trends
in Cognitive Sciences, 16, 200–206. https://doi.org/10.1016/j.
tics.2012.03.002
King, J. R., & Dehaene, S. (2014). A model of subjective report and
objective discrimination as categorical decisions in a vast rep-
resentational space. Philosophical Transactions of the Royal
Society, B: Biological Sciences, 369, 20130204. https://doi.org/
10.1098/rstb.2013.0204
Kloosterman, N. A., De Gee, J. W., Bergner, M. W.,
Lindenberger, U., Garrett, D. D., & Fahrenfort, J. J. (2019).
Humans strategically shift decision bias by flexibly adjusting
sensory evidence accumulation. eLife, 8, e37321. https://doi.
org/10.7554/eLife.37321
Ko, Y., & Lau, H. (2012). A detection theoretic explanation of
blindsight suggests a link between conscious perception and
metacognition. Philosophical Transactions of the Royal Society,
B: Biological Sciences, 367, 1401–1411. https://doi.org/10.1098/
rstb.2011.0380
Koivisto, M., & Grassini, S. (2016). Neural processing around
200 ms after stimulus-onset correlates with subjective visual
awareness. Neuropsychologia, 84, 235–243. https://doi.org/10.
1016/j.neuropsychologia.2016.02.024
Kouider, S., & Dehaene, S. (2007). Levels of processing during non-
conscious perception: A critical review of visual masking. Phil-
osophical Transactions of the Royal Society, B: Biological Sci-
ences, 362, 857–875. https://doi.org/10.1098/rstb.2007.2093
Kouider, S., Dehaene, S., Jobert, A., & Le Bihan, D. (2007). Cerebral
bases of subliminal and supraliminal priming during reading.
Cerebral Cortex, 17, 2019–2029. https://doi.org/10.1093/cercor/
bhl110
Lamme, V. A. F., & Roelfsema, P. R. (2000). The distinct modes of
vision offered by feedforward and recurrent processing. Trends
in Neurosciences, 23, 571–579. https://doi.org/10.1016/S0166-
2236(00)01657-X
Lamy, D., Salti, M., & Bar-Haim, Y. (2009). Neural correlates of sub-
jective awareness and unconscious processing: An ERP study.
Journal of Cognitive Neuroscience, 21, 1435–1446. https://doi.
org/10.1162/jocn.2009.21064
Lau, H. C., & Passingham, R. E. (2006). Relative blindsight in nor-
mal observers and the neural correlate of visual consciousness.
Proceedings of the National Academy of Sciences of the United
States of America, 103, 18763–18768. https://doi.org/10.1073/
pnas.0607716103
Lau, H. C., & Passingham, R. E. (2007). Unconscious activation of
the cognitive control system in the human prefrontal cortex.
Journal of Neuroscience, 27, 5805–5811. https://doi.org/10.
1523/JNEUROSCI.4335-06.2007
Limbach, K., & Corballis, P. M. (2016). Prestimulus alpha power
influences response criterion in a detection task. Psychophysi-
ology, 53, 1154–1164. https://doi.org/10.1111/psyp.12666
Lloyd, D. A., Abrahamyan, A., & Harris, J. A. (2013). Brain-
stimulation induced blindsight: Unconscious vision or
response bias? PLoS ONE, 8, e82828. https://doi.org/10.1371/
journal.pone.0082828
Mäkinen, V., Tiitinen, H., & May, P. (2005). Auditory event-related
responses are generated independently of ongoing brain activ-
ity. NeuroImage, 24, 961–968. https://doi.org/10.1016/j.
neuroimage.2004.10.020
Maris, E., & Oostenveld, R. (2007). Nonparametric statistical test-
ing of EEG- and MEG-data. Journal of Neuroscience
Methods, 164, 177–190. https://doi.org/10.1016/j.jneumeth.
2007.03.024
Mathewson, K. E., Gratton, G., Fabiani, M., Beck, D. M., & Ro, T.
(2009). To see or not to see: Prestimulus α phase predicts visual
awareness. Journal of Neuroscience, 29, 2725–2732. https://doi.
org/10.1523/JNEUROSCI.3963-08.2009
Mazaheri, A., & Jensen, O. (2008). Asymmetric amplitude modula-
tions of brain oscillations generate slow evoked responses.
Journal of Neuroscience, 28, 7781–7787. https://doi.org/10.
1523/JNEUROSCI.1631-08.2008
Nikulin, V. V., Linkenkaer-Hansen, K., Nolte, G., Lemm, S.,
Müller, K. R., Ilmoniemi, R. J., & Curio, G. (2007). A novel
mechanism for evoked responses in the human brain.
European Journal of Neuroscience, 25, 3146–3154. https://doi.
org/10.1111/j.1460-9568.2007.05553.x
Overgaard, M. (2011). Visual experience and blindsight: A method-
ological review. Experimental Brain Research, 209, 473–479.
https://doi.org/10.1007/s00221-011-2578-2
Overgaard, M., Rote, J., Mouridsen, K., & Ramsøy, T. Z. (2006). Is
conscious perception gradual or dichotomous? A comparison
of report methodologies during a visual task. Consciousness
and Cognition, 15, 700–708. https://doi.org/10.1016/j.concog.
2006.04.002
Pernet, C. R., Chauveau, N., Gaspar, C., & Rousselet, G. A. (2011).
LIMO EEG: A toolbox for hierarchical linear modeling of elec-
troencephalographic data. Computational Intelligence and Neu-
roscience, 2011, 1–11. https://doi.org/10.1155/2011/831409
Persaud, N., Davidson, M., Maniscalco, B., Mobbs, D.,
Passingham, R. E., Cowey, A., & Lau, H. (2011). Awareness-
related activity in prefrontal and parietal cortices in blindsight
reflects more than superior visual performance. NeuroImage,
58, 605–611. https://doi.org/10.1016/j.neuroimage.2011.06.081
Peters, M. A. K., & Lau, H. (2015). Human observers have optimal
introspective access to perceptual processes even for visually
4998 RAILO ET AL.
masked stimuli. eLife, 4, e09651. https://doi.org/10.7554/eLife.
09651
Phillips, I. (2018). Unconscious perception reconsidered. Analytic
Philosophy., 59, 471–514. https://doi.org/10.1111/phib.12135
Phillips, I. (2020). Blindsight is qualitatively degraded conscious
vision. Psychological Review, 128, 558–584. https://doi.org/10.
1037/rev0000254
Polich, J. (2007). Updating P300: An integrative theory of P3a and
P3b. Clinical Neurophysiology, 118, 2128–2148. https://doi.org/
10.1016/j.clinph.2007.04.019
Railo, H., & Hurme, M. (2021). Is the primary visual cortex neces-
sary for blindsight-like behavior? Review of transcranial mag-
netic stimulation studies in neurologically healthy individuals.
Neuroscience and Biobehavioral Reviews, 127, 353–364.
Railo, H., Koivisto, M., & Revonsuo, A. (2011). Tracking the pro-
cesses behind conscious perception: A review of event-related
potential correlates of visual consciousness. Consciousness and
Cognition, 20, 972–983. https://doi.org/10.1016/j.concog.2011.
03.019
Rajananda, S., Zhu, J., & Peters, M. A. K. (2020). Normal observers
show no evidence for blindsight in facial emotion perception.
Neuroscience of Consciousness., 2020, niaa023. https://doi.org/
10.1093/nc/niaa023
Ramsøy, T. Z., & Overgaard, M. (2004). Introspection and sublimi-
nal perception. Phenomenology and the Cognitive Sciences, 3,
1–23. https://doi.org/10.1023/b:phen.0000041900.30172.e8
Salti, M., Bar-Haim, Y., & Lamy, D. (2012). The P3 component of
the ERP reflects conscious perception, not confidence. Con-
sciousness and Cognition, 21, 961–968. https://doi.org/10.1016/
j.concog.2012.01.012
Salti, M., Monto, S., Charles, L., King, J. R., Parkkonen, L., &
Dehaene, S. (2015). Distinct cortical codes and temporal
dynamics for conscious and unconscious percepts. eLife, 4,
e05652. https://doi.org/10.7554/eLife.05652
Samaha, J., Iemi, L., Haegens, S., & Busch, N. A. (2020). Spontane-
ous brain oscillations and perceptual decision-making. Trends
in Cognitive Sciences, 24, 639–653. https://doi.org/10.1016/j.
tics.2020.05.004
Samaha, J., Iemi, L., & Postle, B. R. (2017). Prestimulus alpha-band
power biases visual discrimination confidence, but not accu-
racy. Consciousness and Cognition, 54, 47–55. https://doi.org/
10.1016/j.concog.2017.02.005
Sandberg, K., Timmermans, B., Overgaard, M., & Cleeremans, A.
(2010). Measuring consciousness: Is one measure better than
the other? Consciousness and Cognition, 19, 1069–1078.
https://doi.org/10.1016/j.concog.2009.12.013
Schaworonkow, N., Blythe, D. A. J., Kegeles, J., Curio, G., &
Nikulin, V. V. (2015). Power-law dynamics in neuronal and
behavioral data introduce spurious correlations. Human Brain
Mapping, 36(8), 2901–2914. https://doi.org/10.1002/hbm.22816
Schroeder, C. E., & Lakatos, P. (2009). Low-frequency neuronal
oscillations as instruments of sensory selection. Trends in Neu-
rosciences, 32, 9–18. https://doi.org/10.1016/j.tins.2008.09.012
Shah, A. S., Bressler, S. L., Knuth, K. H., Ding, M., Mehta, A. D.,
Ulbert, I., & Schroeder, C. E. (2004). Neural dynamics and the
fundamental mechanisms of event-related brain potentials.
Cerebral Cortex, 14, 476–483. https://doi.org/10.1093/cercor/
bhh009
Siedlecka, M., Hobot, J., Skora, Z., Paulewicz, B.,
Timmermans, B., & Wierzchon, M. (2019). Motor response
influences perceptual awareness judgements. Consciousness
and Cognition, 75, 102804. https://doi.org/10.1016/j.concog.
2019.102804
Smith, S. M., & Nichols, T. E. (2009). Threshold-free cluster
enhancement: Addressing problems of smoothing, threshold
dependence and localisation in cluster inference. NeuroImage,
44, 83–98. https://doi.org/10.1016/j.neuroimage.2008.03.061
Stein, T., & Peelen, M. V. (2021). Dissociating conscious and uncon-
scious influences on visual detection effects. Nature Human
Behaviour, 5, 612–624. https://doi.org/10.1038/s41562-020-
01004-5
Supèr, H., Spekreijse, H., & Lamme, V. A. F. (2001). Two distinct
modes of sensory processing observed in monkey primary
visual cortex (VI). Nature Neuroscience, 4, 304–310. https://doi.
org/10.1038/85170
Thut, G., Nietzel, A., Brandt, S. A., & Pascual-Leone, A. (2006).
α-Band electroencephalographic activity over occipital cortex
indexes visuospatial attention bias and predicts visual target
detection. Journal of Neuroscience, 26, 9494–9502. https://doi.
org/10.1523/JNEUROSCI.0875-06.2006
Van Gaal, S., Ridderinkhof, K. R., Scholte, H. S., & Lamme, V. A. F.
(2010). Unconscious activation of the prefrontal no-go net-
work. Journal of Neuroscience, 30, 4143–4150. https://doi.org/
10.1523/JNEUROSCI.2992-09.2010
Vanrullen, R. (2007). The power of the feed-forward sweep.
Advances in Cognitive Psychology, 3, 167–176. https://doi.org/
10.2478/v10053-008-0022-3
VanRullen, R., & Koch, C. (2003). Visual selective behavior can be
triggered by a feed-forward process. Journal of Cognitive Neu-
roscience, 15, 209–217. https://doi.org/10.1162/0898929033
21208141
Watson, A. B., & Pelli, D. G. (1983). Quest: A Bayesian adaptive psy-
chometric method. Perception & Psychophysics, 33, 113–120.
https://doi.org/10.3758/BF03202828
Weiskrantz, L., Warrington, E. K., Sanders, M. D., & Marshall, J.
(1974). Visual capacity in the hemianopic field following a
restricted occipital ablation. Brain, 97(4), 709–728. https://doi.
org/10.1093/brain/97.4.709
Worden, M. S., Foxe, J. J., Wang, N., & Simpson, G. V. (2000).
Anticipatory biasing of visuospatial attention indexed by reti-
notopically specific α-band electroencephalography increases
over occipital cortex. Journal of Neuroscience, 20, RC63.
https://doi.org/10.1523/jneurosci.20-06-j0002.2000
SUPPORTING INFORMATION
Additional supporting information may be found online
in the Supporting Information section at the end of this
article.
How to cite this article: Railo, H., Piccin, R., &
Lukasik, K. M. (2021). Subliminal perception is
continuous with conscious vision and can be
predicted from prestimulus
electroencephalographic activity. European Journal
of Neuroscience, 54(3), 4985–4999. https://doi.org/
10.1111/ejn.15354
RAILO ET AL. 4999