Consciousness and Cognition 35 (2015) 206–224
Contents lists available at ScienceDirect
Consciousness and Cognition
journal homepage: www.elsevier.com/locate/concog
Ignition’s glow: Ultra-fast spread of global cortical activity
accompanying local ‘‘ignitions’’ in visual cortex during
conscious visual perception
N. Noy a,b, S. Bickel c, E. Zion-Golumbic d,e, M. Harel b, T. Golan f, I. Davidesco f, C.A. Schevon g,
G.M. McKhann g, R.R. Goodman g, C.E. Schroeder d,e, A.D. Mehta c, R. Malach b,⇑
a
Gonda Brain Research Center, Bar-Ilan University, Ramat Gan, Israel
Department of Neurobiology, Weizmann Institute of Science, Rehovot, Israel
Department of Neurosurgery, Hofstra North Shore LIJ School of Medicine, Hempstead, NY, USA
d
Department of Psychiatry, Columbia University College of Physicians and Surgeons, New York, NY, USA
e
Cognitive Neuroscience and Schizophrenia Program, Nathan Kline Institute, Orangeburg, NY, USA
f
The Edmond & Lily Safra Center for Brain Sciences, Interdisciplinary Center for Neural Computation, Hebrew University, Jerusalem, Israel
g
Department of Neurology, Columbia University College of Physicians and Surgeons, New York, NY, USA
b
c
a r t i c l e
i n f o
Article history:
Received 3 November 2014
Revised 5 March 2015
Accepted 6 March 2015
Available online 29 March 2015
Keywords:
Consciousness
Visual awareness
Ignition
Visual cortex
ECoG
Global work-space
Subjective awareness
a b s t r a c t
Despite extensive research, the spatiotemporal span of neuronal activations associated
with the emergence of a conscious percept is still debated. The debate can be formulated
in the context of local vs. global models, emphasizing local activity in visual cortex vs. a
global fronto-parietal ‘‘workspace’’ as the key mechanisms of conscious visual perception.
These alternative models lead to differential predictions with regard to the precise magnitude, timing and anatomical spread of neuronal activity during conscious perception. Here
we aimed to test a specific aspect of these predictions in which local and global models
appear to differ – namely the extent to which fronto-parietal regions modulate their activity during task performance under similar perceptual states. So far the main experimental
results relevant to this debate have been obtained from non-invasive methods and led to
conflicting interpretations. Here we examined these alternative predictions through
large-scale intracranial measurements (Electrocorticogram – ECoG) in 43 patients and
4445 recording sites. Both ERP and broadband high frequency (50–150 Hz – BHF)
responses were examined through the entire cortex during a simple 1-back visual recognition memory task. Our results reveal short latency intense visual responses, localized first
in early visual cortex followed (at 200 ms) by higher order visual areas, but failed to show
significant delayed (300 ms) visual activations. By contrast, oddball image repeat events,
linked to overt motor responses, were associated with a significant increase in a delayed
(300 ms) peak of BHF power in fronto-parietal cortex. Comparing BHF responses with
ERP revealed an additional peak in the ERP response – having a similar latency to the
well-studied P3 scalp EEG response. Posterior and temporal regions demonstrated robust
visual category selectivity. An unexpected observation was that high-order visual cortex
responses were essentially concurrent (at 200 ms) with an ultra-fast spread of signals
of lower magnitude that invaded selected sites throughout fronto-parietal cortical areas.
Our results are compatible with local models in demonstrating a clear task-dependence
of the 300 ms fronto-parietal activation. However, they also reveal a more global
⇑ Corresponding author at: Weizmann Institute of Science, Rehovot 76100, Israel.
E-mail address: rafi.malach@gmail.com (R. Malach).
http://dx.doi.org/10.1016/j.concog.2015.03.006
1053-8100/Ó 2015 Elsevier Inc. All rights reserved.
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
207
component of low-magnitude and poor content selectivity that rapidly spreads into frontoparietal sites. The precise functional role of this global ‘‘glow’’ remains to be elucidated.
Ó 2015 Elsevier Inc. All rights reserved.
1. Introduction
The neuronal events associated with the emergence of a content-specific visual image in human consciousness are still
poorly understood and a source of an ongoing debate. A central question in this research concerns the minimal group of neurons whose activity is both necessary and sufficient for visual awareness and the magnitude and dynamics of this activity.
Despite extensive research this question is far from resolved.
A useful way to frame this issue more concretely is from the perspective of global vs. local models of visual perception.
Local models point to activity confined to content-selective visual neurons as the central mechanism that gives rise to a conscious visual percept (Block, 2005, 2007; Fisch et al., 2009; Malach, 2007; Zeki, 2001, 2003; Zeki & Bartels, 1999). In contrast,
according to global models, a wide-spread ‘‘global work-space’’ (GWS) activation, encompassing fronto-parietal cortex as
well, is the critical mechanism (Baars, 1997; Dehaene & Changeux, 2011; Del Cul, Baillet, & Dehaene, 2007; Sergent,
Baillet, & Dehaene, 2005).
Unfortunately the global vs. local formulation is still rather vague, making it difficult to formulate unequivocal predictions
for specific experimental designs. We therefore present here two extreme alternatives, aimed to illustrate, in a highly schematic form, a specific instance where the local and global models make clearly distinguishable and testable predictions.
The two alternatives are depicted in Fig. 1 which outlines central neuronal activation events that are predicted under the
local vs. global alternatives. Note that these are highly idealized illustrations – aimed to highlight the extreme end-points of
the local–global spectrum. The figure illustrates the contrast in the predictions of the two models in regard to the difference
between passive viewing conditions – in which no overt introspective or motor act is required (illustrated by an eye icon in
Fig. 1) and active viewing conditions – in which participants are asked to explicitly decide and report the content of the
visual stimulus (illustrated by a button press icon).
The common framework, based on ample EEG data, underlying both local and global models, entails a three stage process
that is initiated when a person perceives a visual image (see, e.g., Del Cul et al., 2007). These stages include early bursts of
high activity occurring at 100 ms and associated with early, retinotopic, visual cortex activations (EVA, light blue1 curves in
Fig. 1). The activations are then followed by a high activity burst at 200 ms localized to high order, content-specific visual areas
(HVA, dark blue curves in Fig. 1). Finally, there is a third peak of activity, occurring at 300 ms. This high activity peak has been
termed ‘‘global ignition’’ (Del Cul et al., 2007) and is suggested to engage a wide-spread global network of fronto-parietal cortical areas – the so called ‘‘global work space’’ (GWS, Baars, 1997, 2002; Dehaene, Changeux, Naccache, Sackur, & Sergent, 2006).
Within this common framework the local and global models differ in two critical aspects. First, local models do not predict a mandatory late (300 ms) fronto-parietal activation during the passive viewing condition, while global models predict
a high magnitude and wide-spread activation during this time. This is not a mere technical point, the difference stems from a
fundamental disagreement in assumptions. While according to local models, neural correlates of visual perception per se
begin and end in the visual cortex, the global models’ view is that the visual cortex activation is just a preparatory stage,
and a second, separate, global event is necessary for a conscious visual percept to emerge.
It is important to clarify at this point that local models by no means preclude the possibility of fronto-parietal activation
during passive viewing conditions; they only argue that this activity is not mandatory for the emergence of the contentspecific visual percept. Thus, one can envision a quite likely scenario in which even when perceiving a visual stimulus under
passive conditions, such percepts may elicit, even involuntarily, non-visual cognitive events such as working memory encoding, naming and other linguistic associations and attentional effects. Such processes may inevitably activate non visual
representations, even under the strictest passive viewing conditions.
However, according to local models, these activations reflect a different subjective content from that of the visual percepts and should be viewed as consequences of it. This, potentially tight, link between visual perception and subsequent processes poses a serious methodological challenge when attempting to disentangle the representations of visual content from
these post-processing stages (see Malach, 2007). Fortunately, as will be shown below, there are certain experimental conditions in which activations related to visual perception can be disentangled from those associated with post-perceptual cognitive events.
A second difference between local and global models concerns the change in signals associated with the addition of a
report component to the visual percept. Local models predict a significant increase of activation in the global 300 ms component when moving from passive viewing to active report, while global models do not predict additional large scale
increases. This difference stems from the assumption of local models that the subsequent (300 ms) fronto-parietal activation
does not reflect the visual percept itself, but rather subsequent non-visual aspects such as semantic processing or motor preparation and hence should be selectively activated when such post-visual processes are recruited during active report.
1
For interpretation of color in Fig. 1, the reader is referred to the web version of this article.
208
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
Fig. 1. Predictions of neuronal activation profiles in the ‘‘Local’’ and ‘‘Global’’ alternatives. The models are presented in the framework of three peaks of
activations: an early (100 ms post stimulus presentation) peak localized to early visual areas (EVA), an intermediate peak (200 ms) localized to high
order visual areas (HVA) and finally a late peak (300 ms) localized to the global work space at the fronto-parietal cortex (GWS). The different models
predict different profile of activations when comparing passive with active (that demand report) viewing. Top left panel – activity profile predicted by the
local model during passive viewing. Note the lack of activation of the late fronto-parietal component, associated in local models with report rather than
perception. Bottom left panel – activity profile predicted by the local model during active viewing. Here the model predicts high activity of the late
component in line with its post-perceptual specialization. Top right panel – activity profile predicted by the global model during passive viewing. Note the
high activation of the 300 ms component – in line with the prediction of global models that the late, fronto-parietal component is part of the perception
related processes. Bottom right panel – activity profile predicted by the global model during active viewing (report). Since the 300 ms component in the
GWS is fully activated during conscious perception it is not further activated during report. As can be seen, manipulating the task-demand, i.e. from passive
to active viewing, leads to different and testable predictions – i.e. high modulation in local models, and low modulations in global ones.
In contrast, global models assume that the late, high activation in the GWS is already part of the mechanism eliciting the
visual percept itself so they do not expect substantial further activation, at least within the 300 ms window, when a report is
added to the visual percept. However, it should be emphasized that the global alternative does not preclude task-related
activations in additional, localized, cortical areas such as motor cortex which are beyond the GWS networks associated with
conscious perception proper.
Thus, examination of neuronal activations across the cortex, during the viewing of visual images while manipulating the
active report component, could provide evidence that supports one or the other model. Furthermore, this examination could
differentiate those fronto-parietal regions that are activated during passive viewing from those that are specifically taskrelated.
An important point to emphasize with regard to Fig. 1 is that the different predictions of the models do not rely on manipulation of target awareness. In both cases the images are presented well above the awareness threshold and are consciously
perceived. This point is important since a common misconception is that the only means to explore models of conscious perception require manipulations of the target-awareness level under identical physical stimulation. While such elegant manipulations have indeed provided important insights regarding the identification of neuronal mechanisms of content-specific
perception (e.g. Del Cul et al., 2007; Fisch et al., 2009; Frässle, Sommer, Jansen, Naber, & Einhäuser, 2014), they are certainly
not the exclusive means to gain important information with regards to global vs. local alternatives (as explained above).
A large body of experimental data relevant to these fundamental issues has been obtained mainly from noninvasive EEG
and MEG recordings. Unfortunately, this research has led to contrasting views concerning the specific EEG wave that should
be linked with visual awareness.
In an important body of research the Dehaene group has proposed that a late, 300 ms wave (also referred to as P3) is a
‘‘signature’’ of conscious perception in the human brain (Del Cul et al., 2007). Support from EEG and ECoG studies for a global
signal involvement has been provided by other studies as well (Gaillard et al., 2009; Kouider et al., 2013; Lamy, Salti, & BarHaim, 2009; Pins & Ffytche, 2003).
However, a contrasting view has been put forward by Pitts, Metzler, and Hillyard (2014) pointing to a 200 ms negative
wave, likely confined to visual cortex, as the dominant awareness related event (see also review by Koivisto & Revonsuo,
2010). Similarly, intracranial recordings of visual responses in the visual cortex emphasize a shorter latency response in high
order visual areas, occurring approximately at 200 ms latency (Fisch et al., 2009). However, due to the rarity of such intracranial recordings, the number and coverage of the electrodes in this study were rather limited.
Here we tested the predictions presented in Fig. 1 using a particularly large set of intracranial ECoG recordings encompassing thousands of contacts. Such approach offers both high temporal resolution, good cortical localization (Privman et al.,
2007) and extensive coverage of cortical areas. Our study consisted of a large cohort of patients (43) all of which performed
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
209
the same visual (1-back memory) task, providing a uniquely extensive number of cortical recording sites (4445) placed over
all cortical lobes.
Our results are compatible with the local models prediction in that they show that the appearance of the third 300 ms
component was associated mainly with the target condition, involving report. However, a surprising finding, not predicted
by either alternative has been the discovery of low magnitude signals that spread ultra-rapidly (overlapping in time with the
200 ms wave of activation in the visual cortex) to a large number of specific sites located across all cortical lobes.
Thus, our results suggest a ‘‘hybrid’’ model composed of an intense local activation in high order visual areas, concurrent
in time with a global, non-specific and of lower magnitude component that engages many fronto-parietal sites.
2. Methods
2.1. Participants
Recordings of electrical activity were obtained from 48 neurosurgical patients (20 female, aged 32 ± 10.7, mean ± SD),
with pharmacoresistant epilepsy, monitored for potential surgical treatment (see Table 1). Of all patients, 40 were hospitalized at the North Shore LIJ Hospitals and eight at Columbia University Medical Center. Four patients with extremely low task
performance (no hits at all, see Section 2.6 and Table 1) were excluded from analysis together with one patient with highly
corrupted signals, leaving 43 patients for analysis. All patients provided fully informed consent according to the National
Institutes of Health guidelines, as monitored by the local institutional review board, in accordance with the ethical standards
of the Declaration of Helsinki. The experimental protocol and study execution were approved by the Sourasky Medical
Hospital and local institutional review boards.
ECoG recordings were made over the course of clinical monitoring for spontaneous seizures. The decision to implant, the
electrode location, and the duration of implantation were determined entirely based on clinical grounds without reference to
this investigation. The patients were informed that participation in this study would not jeopardize their clinical care in any
way and that they could withdraw from it at any time. In fact, it should be noted that participating in these studies may have
actually benefitted the patients by closer monitoring and multiple observer and scientific personal confirmation of electrode
identities and localization.
2.2. Stimuli and task
The patients went through a one back visual memory task (see Fig. 2, Privman et al., 2007) in which images from four
categories were presented pseudo randomly at a rate of one image per second – each presented for 250 ms and followed
by a gray screen for an interval of 750 ms. The patients were requested to press a computer mouse button whenever two
consecutive pictures were identical and to refrain from pressing in any other case. They were instructed to fixate on a small
point superimposed on the images and blank intervals and to respond as accurately as possible within the time frame of a
trial, no encouragement was given in regard to reaction time.
The set of images included 40 different exemplars divided into four categories: 15 faces, 10 houses, 10 man-made objects
and 5 patterns. Only exact exemplar repetitions were counted as targets, category repeats (about 25%) are non-targets by
definition. The images were in gray scale 12° (500 pixels) in width. All were presented on a standard laptop display
(60 Hz refresh rate) at a distance of about 60 cm. Patients’ responses were recorded via a mouse button press.
Each patient underwent one or two runs of the experiment consisting a total of 121–432 trials, of which about 12% were
target images (identical to the image preceding them).
2.3. Data acquisition
Each patient was implanted with subdural electrode arrays containing 52–207, two millimeter in diameter, contact electrodes (see Table 1). In the subset of 43 patients included in the current analysis the electrode number ranged from 52 to 200,
summing up to a total of 5218. Electrodes were arranged in one-dimensional strips or in two dimensional grids placed
directly on the cortical surface.
The signal was sampled at a rate of 2000 Hz (North Shore University Hospital) or 1000 Hz (Columbia). Stimulus-triggered
electrical pulses were recorded along with the ECoG data for precise synchronization of the stimuli with the electrical
responses.
2.4. Electrodes localization
Computed tomography (CT) scans following electrode implantation were co-registered to the postoperative MRI. Pre and
postoperative MRIs were both skull-stripped using the BET algorithm from the Oxford Centre for Functional MRI of the Brain
(FMRIB) software library (FSL; www.fmrib.ox.ac.uk/fsl/), followed by coregistration to account for possible brain shifts
caused by electrode implantation and surgery. Electrodes were identified in the CT image using BioImagesuite (www.bioimagesuite.org). The coordinates of the electrodes were then normalized to Talairach space (Talairach & Tournoux, 1988),
210
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
Table 1
Patients’ characteristics.
In gray – patients that were excluded from the analysis.
F = female, M = male, O = occipital, F = frontal, T = temporal, P = parietal, FP = false positive, RT = reaction time.
Fig. 2. Experimental paradigm – oddball, one back visual memory task. In each trial an image (face, house, object or pattern) was presented for 250 ms,
followed by a fixation only interval of 750 ms. The patients were requested to press a mouse button in a minority (12%) of trials when an image was
identical to the image preceding it (target) or, in most of the trials, refrain from pressing when the image was different (non-target).
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
211
enabling precise localization both with relation to the patients’ anatomical MRI scan and in standard coordinate space. This
process left out 44 electrodes which were undetectable in the CT and their coordinates could not be extracted, and another
set of 101 electrodes that were located in lesion areas or their proximity.
For joint presentation of all patients’ electrodes, electrode locations were projected onto a cortical reconstruction of a
healthy subject from a previous study. The electrodes were then divided into seven functional groups, corresponding to
rough anatomical segregation: low order visual (LOV), high order visual (HOV), parietal (Prt), auditory (Aud), post central
(PoC), precentral (PrC) and frontal (Frt) electrodes. This division left out 283 electrodes that could not be attributed to
any of the regions according to their Talairach location. These electrodes, together with the 145 electrodes mentioned above
and 345 electrodes with corrupted signal were excluded from the analysis, leaving us with 4445 electrodes.
2.5. Data preprocessing
First, all signals were down sampled to 500 Hz. Then, for each patient, a ‘‘common’’ ECoG time course was calculated by
averaging the raw time courses from all of his electrodes (only those that were included in the analysis). Each electrode was
re-referenced by subtraction of this common component, thus discarding non-neuronal contributions from the extracranial
reference electrode (Fisch et al., 2009; Meshulam et al., 2013; Privman et al., 2007; Ramot et al., 2012).
For the ERP we used the absolute value of a 1–25 Hz bandpass filtered signal.
For the calculation of the broadband high gamma frequencies (BHF) power modulations, the signal was first band passed
in the frequency range 48–154 Hz using linear-phase FIR filters (window size ranging from 30 to 60 ms according to the subrange – see next paragraph). Then, the power modulations were extracted by taking the absolute value of the Hilbert transform of the filtered signal (Fisch et al., 2009; Lachaux et al., 2005).
To offset the roughly 1/f2 profile of the power spectrum, which results in dominant contributions from the lower frequency end of the spectrum, we calculated a ‘‘flattened’’ form of the BHF by dividing the entire frequency range into nine
sub-ranges of 11 Hz width, calculating and scaling the power modulation individually in each sub range by its mean value,
and finally averaging across sub-ranges (Fisch et al., 2009; Vidal et al., 2010). In this process we excluded the AC frequencies
from the calculation (the sub-ranges did not include 59–61 Hz and 119–121 Hz).
Data processing was carried out using MATLAB. For filtering, we used original and adapted EEGLAB code (Delorme &
Makeig, 2004).
2.6. Behavioral task performance
The patients were requested to press a computer-mouse button whenever two consecutive pictures were identical, and to
refrain from pressing in any other case. The percent of correct button presses and the mean RT for these trials were assessed
for each patient.
The correlation between the percent of correct responses and the average magnitude of neuronal responses (see
Section 2.7) across patient was calculated per region to assess the impact of behavioral performance on the measured brain
activity.
2.7. Quantitative definitions of electrode responses
To examine the responses to picture presentation under different task demands, we defined two kinds of trials: target
trials, in which a motor report was required (a presentation of a picture that was identical to the one preceding it), and
non-target trials, in which no action was requested (presentation of a different picture than the preceding). In the two trial
types we included only correct responses; errors (misses and false alarms) were omitted from the analysis since their underlying cognitive process is more difficult to interpret.
For each electrode separately, single trials’ BHF responses (0–750 ms post stimulus presentation) were compared to their
baseline intervals (250 to 0 ms). The comparison was based on the areas under the curves of these intervals’ time courses
(AUC, defined here as the mean of the interval’s time point values). Electrodes were defined as responsive if their response
AUCs were significantly larger compared to their baseline AUCs over trials (Davidesco et al., 2013b).
The statistical significance was determined by means of one sample one tailed Student’s t test (trials are degrees of freedom). P-value was calculated for each of the 4445 electrodes examined in this study and then corrected for multiple comparisons to a .05 false discovery rate – FDR (Benjamini & Hochberg, 1995). This process was done once with the non-target
trials and once with the target trials, defining two sets of electrodes. The first set (showing significant response during nontarget trials) will be referred to as NT electrodes and the second (showing significant response during target trials) as T
electrodes.
The magnitude and latency of response were calculated for the target and non-target conditions separately, in each electrode (both for ERP and for BHF). The magnitude of response is the mean of the AUCs over trials, each normalized by its baseline to represent average percent signal change. The latency was defined as the time point where the signal, averaged over
trials, reached 1n4 of its maximum value (within the post stimulus onset window). It should be noted that this criterion is
conservative; it leads to an over-estimate of the true latency expected under ideal SNR conditions.
212
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
The magnitude values (AUCs) were used to calculate an electrode response modulation index (RMI) – a value which
represents the change in the AUC between target and non-target neural responses. It was calculated as: |(AUCt AUCnt)/
(AUCt + AUCnt)|. Note that this is an absolute value, representing the response modulation regardless of its direction (for
identical AUCs the RMI score is zero).
Finally, in order to assess categorization effects, we compared the AUCs produced in response to each category of visual
stimuli separately. For each electrode, the single trial AUCs were divided by category to 4 groups and the significance of the
difference between them was assessed by a one way ANOVA test (corrected for multiple comparisons with .05 FDR). Each
electrode was assigned the F value of the ANOVA test to represent its category selectivity index (CSI).
Overall, across patients, average values (induced response, AUC, latency, RMI and CSI) for the different brain regions were
produced in two steps: First, we calculated a value across trials for each responsive electrode separately (as explained above)
then we averaged all active electrodes in each region across all patients. Population histograms were calculated to depict the
distribution of indexes of individual electrodes across the population. All values were calculated separately for the BHF signal
and for the filtered 1–25 Hz signal (ERP).
3. Results
3.1. Behavioral responses
The average hit (pressing a button when a second identical image was presented) percentage across all 43 analyzed
patients was 65.1 ± 24.5% (mean ± SD), false alarm rate (pressing a button when non-identical images appeared) was
5.2 ± 8.8% and the average RT in hit trials was 569.9 ± 96.7 ms. Performance levels for individual patients are depicted in
Table 1.
As can be seen, there was a substantial behavioral variability in individual patients’ performance – likely attributable to
the fact that the experiments were conducted shortly following the electrode implantation surgery. Thus, care was taken to
assess whether the behavioral variability had a significant impact on the experimental results across the patient population
(see behavioral correlates in Section 3.3).
3.2. Neuronal responses
As we and others have reported previously (Fisch et al., 2009; Henrie & Shapley, 2005; Mukamel et al., 2005; Nir et al.,
2007; Ray & Maunsell, 2011; Tallon-Baudry & Bertrand, 1999) an important index of averaged neuronal firing is the power in
the broadband high frequency range of the local field potential (BHF, 50–150 Hz, also known as gamma). We therefore
focused our analysis on this frequency band. To allow better comparison to scalp EEG recordings we also calculated the
ERP response. However, in order to avoid signal cancellations due to local dipole inversions when calculating group
responses, we have used the absolute value of the ERP (see Section 2.5).
To examine the dynamics of the visually induced BHF responses we averaged the power modulations across individual
trials, time locked to the visual stimulus onset. Target and non-target trials were analyzed separately. It should be emphasized that covert motor responses or other non-visual processes such as inner speech or memory could not be ruled out also
in the case of non-target trials.
Examining the cortical activation during picture presentations revealed an extensive activation pattern. Fig. 3 depicts the
distribution of the entire set of electrodes presented on inflated (A) and flattened cortical formats (B). Note that, due to the
large number of patients and electrodes, our coverage was substantial and included all cortical lobes. However, the electrodes were placed on the cortical surface so deep sulci could not be sampled (marked as dark gray areas in Fig. 3).
Of the 4445 electrodes that were analyzed, 678 (15.3%) showed a significant response during the non-target trials (NT
electrodes). A different but partially overlapping set of 678 (15.3%) electrodes showed a significant response during the target trials (T electrodes). Throughout the signal analysis below, we will compare the responses in these two sets of electrodes
and the responses to both types of trials in the NT electrodes set.
As can be seen in Fig. 3, and in agreement with previous studies, the NT electrodes were located predominantly in the
expected visual areas, while the T electrodes were found mainly in non-visual cortex. However, scattered electrodes violating this tendency could be observed as well. Thus, significant visual responses could be discerned also outside visual cortex,
with prominent clusters in the parietal and frontal lobes (unfilled arrowheads). The part of the T electrodes that showed significant responses only during the target condition were likely activated either by the oddball targeted image or the explicit
motor (button press) response rather than by the visual images presented during the trials.
Given the large number of electrodes and in order to facilitate the data presentation, we subdivided the electrodes into
groups based on rough anatomical considerations (see color contours in Fig. 3B). It should be emphasized that these subdivisions were approximate since independent functional delineation based on fMRI was unavailable in most of the patients.
Three main aspects were discerned in the results. The first was the latency of the responses across different parts of the
brain and different task demands. The second was the relative magnitude of responses across different cortical regions. The
third was the functional selectivity of the responses, i.e. to what extent the activation was modulated by the task (target or
non-target trials) or by the stimulus category – represented by RMI and CSI respectively (see Section 2.7).
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
213
Fig. 3. Location of ECoG cortical recording sites in all patients. Sites were obtained from CT and MRI scans, superimposed on a cortical reconstruction of one
subject from a previous study. (A) The electrodes are shown on a lateral view of the cortical hemispheres. They are color coded according to the condition
they responded significantly to (non-target, target or both). (B) Same electrodes, with the same color coding as in panel A are shown on a flattened map of
the brain. Filled arrowheads indicate clusters of significant electrodes in occipital and temporal lobes and unfilled arrowheads indicate clusters in frontal
and parietal lobes. Color lines represent the subdivision of electrodes according to functional areas. LOV, low order visual; HOV, high order visual; Prt,
parietal; Aud, auditory; PoC, postcentral; PrC, precentral; Frt, frontal; P, posterior; A, anterior; LH, left hemisphere; RH, right hemisphere. (For interpretation
of the references to color in this figure legend, the reader is referred to the web version of this article.)
3.2.1. Latency of response
A particularly consistent phenomenon, evident throughout the entire set of electrodes and across all brain areas, was the
relatively short latency of visually induced neural responses. In particular, the latency of visual responses in non-visual cortical areas was essentially the same as that of the responses in high order visual cortex. This can be appreciated by comparing
the rise time of the normalized fronto-parietal responses with that of the high order visual areas. The results are depicted in
Fig. 4 showing a comparison between the average BHF responses (across all electrodes) of each area for the two sets of electrodes – NT and T, and for the two types of trials – non-target and target. Fig. 5 shows the same averages in the ERP domain.
Examining the induced responses during target trials revealed that the response in non-visual areas show a prominent
delayed peak that shifted toward long (300–400 ms) latencies likely related to the button press, this was evident in both
NT and T sets of electrodes (Fig. 4E and F). Quantitative summaries of the parameters extracted from the results depicted in
Fig. 4A (non-target trials in NT electrodes) are shown in histogram form in Fig. 6 while Fig. 8 shows the histograms for the
results depicted in Fig. 4B (target trials in NT electrodes). Per area averages of these histograms are shown in Figs. 7 and 9
respectively.
The distribution of latencies of non-target responses in NT electrodes is depicted in Figs. 6A (per region) and 7A (across
regions, see Section 2.4). Note that the large majority of electrodes across all cortical areas showed an onset latency of less
than 200 ms. However, few exceptions could be discerned in frontal and parietal areas where longer latencies were also
evident.
Examining the ERP responses – both the time courses (Fig. 5) as well as individual (Fig. 10) and average distributions
(Fig. 11) revealed similarities to the BHF but also interesting discrepancies. In visual cortex, the ERP responses were
characterized by two peaks rather than the unimodal response observed in the BHF result – one peak was at a latency of
100–150 ms while the second was at a latency of 250–350 ms. The majority of electrodes showed ERP onset latency below
200 ms; however there were a substantial number of exceptions with longer latency. Furthermore, the overall peak latency
of ERP response appeared to be located around 300 ms post stimulus onset (Fig. 5). Examining ERP responses produced from
a signal that was preprocessed with a less stringent filter (40 instead of 25 Hz) revealed identical results.
3.2.2. Magnitude of response
A basic response characteristic that has not been directly compared so far across human cortical areas is response magnitude. Here our large data set allowed for a detailed comparison. We estimated response magnitude by calculating the area
under the curve of the BHF and ERP responses (see Section 2.5). Figs. 6B and 7B show a comparison of the magnitude of BHF
responses in individual electrodes distributions and in average region responses respectively.
As can be seen, a consistent signature of visual cortex electrodes compared to non-visual cortex electrodes was a higher
overall response magnitude. This effect was evident particularly in NT electrodes during non-target trials
(Figs. 4A, 6B and 7B), but also during the target trials (Figs. 4B, 8B and 9B) and even in T electrodes in target trials
214
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
Fig. 4. Averaged BHF responses during non-target and target trials in the cortical subdivisions. Panels A and D show the averages of non-target responses in
NT electrodes, panels B and E show the averages of target responses in NT electrodes and panels C and F show it for target responses in T electrodes. Top
panels (A–C) show the original averaged percent signal change while in the bottom panels (D–F) each region average is normalized to its peak amplitude.
Thick lines indicate average, and transparent areas indicate ±SEM. Red bars on the horizontal axis indicate the duration of the visual stimuli. The event
related averaging of the BHF is locked to the onset of the visual stimuli (vertical lines at time zero). (For interpretation of the references to color in this figure
legend, the reader is referred to the web version of this article.)
Fig. 5. Averaged ERP Responses during non-target and target trials in the cortical subdivisions. Same as Fig. 4 only with absolute ERP responses.
(Fig. 4C). Statistical analysis revealed that under all of these conditions the average response magnitude was significantly
higher in visual cortex electrodes compared to non-visual ones (two sided t-test between the two groups yield p < .001 corrected in each of the tests). Furthermore, inspecting the individual electrode distributions (Fig. 6B) – revealed that only visual
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
215
Fig. 6. Latency, AUC, RMI and CSI distribution of BHF responses in NT electrodes. (A) Distribution of response latency during non-target trials. Note the
similar median latency of HOV area and fronto-parietal regions. (B) Non-target trials AUC distribution. Note the small AUC characterizing fronto-parietal
region as compared to visual regions. (C) Response modulation index (RMI) distribution. (D) Category selectivity index (CSI) distribution. Red plus on the x
axes marks the median of the distributions. AUC, area under the curve; RMI, response modulation index; CSI, category selectivity index; BHF, broadband
high frequency; NT, non-target. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
cortex electrodes manifested high AUC responses during the non-target condition. The ERP measurements demonstrated
similar results (Figs. 10B and 11B).
3.2.3. Response modulation
A third important aspect of the electrodes’ response properties was their functional selectivity. Here our experimental
paradigm allowed us to ask two complementary questions.
First, to what extent ECoG recording sites manifested task-response modulation – i.e. showed a difference between the
non-target trials and the target trials. To quantify this property we calculated a response modulation index (RMI, see
Section 2.7). The spatial distribution of the RMI is depicted in Fig. 12, in which all significantly responsive electrodes (i.e. both
NT and T sets) are color coded according to their RMI score.
Quantitative comparison of the RMI averages and individual electrode distributions in NT electrodes are shown in
Figs. 7C and 6C respectively. As can be seen, there was a clear tendency for response modulation to be dominant in non-visual cortex electrodes. However, clear clusters of non-modulated electrodes could also be discerned in frontal cortex, particularly the inferior frontal gyrus (see Fig. 12). On average, there was a significant preference for response modulated
electrodes to be localized in non-visual cortex regions – the mean RMI in these regions was 0.49 (N = 204) compared to
0.22 (N = 456) in visual areas (p < .001, two sided t-test).
216
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
Fig. 7. Latency, AUC, RMI and CSI area averages of BHF responses in NT electrodes. Averages of the distributions depicted in Fig. 6 – in the x axis the different
brain areas and in the y axis average values (outliers excluded) for latency (A), AUC (B), RMI (C) and CSI (D).
3.2.4. Category selectivity
Finally, we examined to what extent individual ECoG sites manifested visual category selectivity in their response magnitudes. Fig. 13 depicts the distribution of category selectivity index (CSI, see Section 2.7 for calculation) across significantly
selective electrodes. As can be seen, the CSI distribution was broadly the inverse of that of the RMI, with high selectivity confined exclusively to visual cortex electrodes. Few cases of low but significant CSI could be discerned also in frontal regions
electrodes.
The quantitative analysis of average values (Fig. 7D) and individual electrode distributions (Fig. 6D) reflect these differences. Statistical analysis revealed a significant difference between visual (m = 8.53, N = 457) and non-visual (m = 1.79,
N = 205) areas (p < .001, two sided t-test). Interestingly, ERP CSI differences across areas (Fig. 11D) and electrodes
(Fig. 10D) were significantly weaker compared to the BHF responses with visual areas’ mean CSI of 2.2 (N = 457) and
non-visual areas’ mean CSI of 1.08 (N = 204, p < .001, two sided t-test). This effect can be appreciated by inspecting the
map of electrodes distribution in Fig. 14 (where only CSI significant electrodes are shown).
3.3. Behavioral correlates
The sub-optimal and variable performance of the patients in the 1-back task raises the concern that the difference
between non-target and target trials may be related to differences in attention. We used only correct trials in both conditions, but correct target trials necessitated high attention while the level of attention in the non-target trials was hard to
determine.
To assess the impact of performance on our results we examined the correlation between the percent of hits (percent of
correct responses to target images) and the non-target condition AUC average for each brain region over participants. This
revealed no significant effect (all p values were greater than .05, FDR corrected).
In addition, focusing the analysis only on highly successful participants (with a hit rate exceeding 75% – N = 19, m = 90.2%)
and comparing the mean RMI score in visual and non-visual brain regions revealed a similar effect to the one found in the
entire group (see Section 3.2.3): non-visual RMI was 0.35 (N = 93) while visual RMI was 0.17 (N = 217), this effect is also
highly significant (p < .001, two sided t-test).
4. Discussion
4.1. Global perspective
Before discussing the results it is important to emphasize the global perspective of our study. In a work based on 4445
individual recording sites taken from essentially the entire human cortex a major challenge is to carefully select the scope
and focus of the analysis so as not to get swamped by the massive, rich and diverse set of individual responses. Our aim here
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
217
Fig. 8. Latency and AUC distribution of BHF responses during target trials in NT electrodes. Same as panels A and B in Fig. 6 only here for the non-target
responses in the same set of electrodes (NT). Note that the RMI calculation takes into account both types of responses, hence it is shown only in Fig. 6.
Fig. 9. Latency and AUC area averages of target BHF responses in NT electrodes. Averages of the distributions depicted in Fig. 8 – in the x axis the different
brain areas and in the y axis average values (outliers excluded) for latency (A) and AUC (B).
was to specifically focus on the dominant and most consistent functional and dynamic ‘‘themes’’ which will be described
below.
This approach comes at the price of lowering our sensitivity to phenomena that occur more sparsely in small and specialized clusters of electrodes. However, to our mind, such broad perspective is a mandatory first step in understanding the flow
218
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
Fig. 10. Latency, AUC, RMI and CSI distribution of ERP responses during non-target trials in NT electrodes. Same as Fig. 6 only here for ERP responses.
of visual signals across the human cortex. It should also be stressed that such ‘‘smoothing’’ of data invariably occurs in noninvasive methods such as scalp EEG and MEG, so our approach may make it easier to compare the invasive ECoG results with
the more common non-invasive approached.
Finally, it should be noted that the data was obtained from epileptic patients, and thus may be affected by various factors
such as the clinical conditions, effects of surgery, medication, etc. Such effects were indeed evident in the rather poor and
variable performance of the patients (see Table 1). However, we have failed to find a significant relationship between
patients’ behavioral performance and our main reported effects (see Section 3.3).
4.2. Model predictions
Considering the local vs. global alternatives depicted in Fig. 1, our results largely support the former (but see the ‘‘measurement window’’ caveat at Section 4.6). Thus, in the majority of recording sites, and particularly in fronto-parietal ones, the
third, 300 ms component was not evident during the non-target condition. Thus, we failed to see a generalized activation of
fronto-parietal cortex as a separate event following the 200 ms visual cortex activation during non-target trials.
By contrast, a prominent activation at 300 ms appeared during the target trials – both in the BHF and ERP components
(see Figs. 4 and 5). Our results show that the vast majority of electrodes that displayed a strong link to target trials were
localized outside of visual cortex. This is in line with the role of fronto-parietal networks in post-perceptual processes associated with overt task execution.
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
219
Fig. 11. Latency, AUC, RMI and CSI area averages of non-target ERP responses in NT electrodes. Same as Fig. 7 only here for ERP responses.
Fig. 12. Response modulation index (RMI). Colors indicate RMI value (yellow for high, red for low) across the entire set of significantly (either during target
or non-target trials) responsive electrodes, i.e. both T and NT. White dashed line separates the visual areas from other parts of the brain. Note that frontoparietal regions (anterior to the white line) were dominated by task modulated electrodes (yellow) while occipito-temporal, high order visual areas
(posterior to the white line), were characterized by many task-invariant electrodes (red). Gray depicts non-responsive electrodes.
It is important to emphasize here as well that, as we noted in the introduction, the present results do not imply that postperceptual frontal activations are completely precluded from occurring in the absence of an overt task. They do, however,
indicate that the late fronto-parietal activation is not necessary for the emergence of content-specific visual images in
awareness.
However, caution needs to be exercised in interpreting these findings. First, the source of the increased activation associated with the target events has not been examined in detail. For example, since the picture repeats were a rather rare event
(12%) it could be that part of the increased activity was related to the ‘‘surprise’’ or novelty factors of these oddball events.
Furthermore, it is important to clarify that while, in terms of overt motor responses, the non-target condition can be considered visual only, patients were still required to attend the images and make a decision to refrain from pressing the button.
Hence, it is likely that some level of motor-planning, linguistic processing and decision making was present in it as well. Such
post-perceptual processes may underlie the weak responses that were observed in fronto-parietal cortex during non-target
trials in the absence of explicit motor response (see Figs. 4A and 5A).
It could also be argued that the report-modulated frontal activity may have actually reflected visual-perceptual aspects
that were specifically linked to the task performance. For example, it could be the case that patients examined the images
220
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
Fig. 13. BHF category selectivity index (CSI). Colors indicate CSI value (yellow for high, red for low, see Section 2.7) across the set of significantly selective
electrodes. White dashed line separates the visual areas from other parts of the cortex. Note that fronto-parietal regions (anterior to the white line) were
dominated by low selectivity electrodes (red) while occipito-temporal, visual areas (posterior to the white line), were characterized with high selectivity
electrodes (yellow). Gray depicts non-selective electrodes.
Fig. 14. ERP category selectivity index (CSI). Same as Fig. 13 only here for ERP.
more thoroughly or more intently upon repeats, so as to verify whether or not to initiate report. However, we think that the
experimental design as well as the observed results argue against this possibility.
First, it should be noted that the images were presented for a brief duration (250 ms, see Section 2.2) thus precluding a
‘‘second look’’ or long inspection strategy. Furthermore, substantial modulations of attention and scrutiny are expected to
first and foremost affect visual cortex sites, as has been amply demonstrated in previous ECoG recordings (e.g. Davidesco
et al., 2013a). In our case, task-related effects were mainly confined to non-visual areas, and if anything, were also reflected
in lower levels of activation in high order visual cortex. These effects were likely due to image repeats, which have been previously documented to lead to a reduction of signal amplitude in visual areas (Grill-Spector & Malach, 2001).
While most fronto-parietal sites displayed strong task modulation, few local clusters, particularly in inferior-frontal
regions were task invariant (see Fig. 12). These sites may be related to non-overt task responses. An interesting possibility
to consider is that they may be a part of the recently reported ‘‘multiple demand’’ system, found in inferior-frontal cortex
that shows effort-related activations (Fedorenko, Duncan, & Kanwisher, 2013).
4.3. Ultra-fast spread of low magnitude global responses during passive viewing
According to strict localist models of perceptual awareness, the neuronal event that is directly associated only with a content-specific conscious percept should be confined to high order visual areas. Our ECoG recording results are not compatible
with such strict view.
Thus, our recordings reveal that neuronal activations, essentially coinciding with the 200 ms wave, could be detected at
each and every cortical subdivision from which recordings were obtained (see Fig. 3). It is important to emphasize, however,
that these short-latency responses were not distributed uniformly across all electrode sites in non-visual areas, but were
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
221
localized to specific anatomical sites (see Fig. 3). These relatively low magnitude responses, likely emanating from the high
magnitude activations in visual cortex, spread throughout the cortex at ultra-fast speed (the median latency of non-target
response in the non-visual areas was 74-118 ms compared to 110 ms in high order visual cortex, Figs. 4D, 6A and 7A).
ERP measurements from the same electrodes also corroborated the finding of short latency onset in specific cortical sites
across non-visual cortical regions (see Figs. 5A, 10A and 11A); this was evident even when a less stringent filter (40 Hz) was
applied.
Our findings are compatible with Foxe and Simpson (2002) EEG study and with recordings in primates, particularly those
obtained from the frontal-eye fields, showing early onset fronto-parietal responses as well as rapid flow of visual signals into
dorsal and ventral stream visual areas (Mehta, Ulbert, & Schroeder, 2000a, 2000b; Schroeder, Mehta, & Givre, 1998).
It could be argued that the fast spread of signals may be a trivial by-product of signal processing – i.e. it could be that the
process of subtracting the global ‘‘common’’ signal from all electrodes (see Section 2.5) may have introduced inadvertently a
‘‘visual’’ like signal into the fronto-parietal electrode responses. To rule out this possibility we have examined the ‘‘common’’
signals in each patient and analyzed their visual responses. As can be seen in Figs. 4 and 5 (dashed lines) the common signals
failed to show significant visual responses at any latency – so these signals could not contribute in any manner to the robust
visual responses we found outside of visual cortex areas.
Furthermore, such global common removal should have introduced a uniform effect across all non-visual electrodes while
our data (Fig. 3) show that the visual responses were present only in specific electrodes in fronto-parietal regions. Thus we
can safely conclude that the fast visual responses we find in fronto-parietal regions are a genuine and robust phenomenon
that is localized to the recording sites proper and cannot be attributed to global sources.
Thus, our results confirm that the cortical network properties are sufficient to support ultra-rapid spread of visual information even across large cortical distances. The routes by which the fronto-parietal responses are generated, whether rapid
flow from the visual cortex (Lamme & Roelfsema, 2000) or indirectly via thalamic loops (Sherman & Guillery, 1996) remains
to be experimentally established in the human brain.
What could be the functional role of the low magnitude global ‘‘glow’’? An important aspect of this activity, which should
be considered when interpreting this effect, is its content invariance. Thus, in contrast to the prominent 200 ms activation in
high order visual areas, which show a clear selectivity effect to the content of the visual images, the global ‘‘glow’’ was largely
content invariant (see Figs. 13 and 14).
It should be noted that we did not manipulate the spatial location of the visual images, so at present we cannot rule out
the possibility that the global glow may have carried spatial location information that could guide attentional resources or
oculomotor responses. Furthermore, it should also be pointed out that while our results show a significant reduction in content selectivity of fronto-parietal responses, they do not rule out the possibility that more sensitive approaches such as
multivariate analysis may nevertheless reveal residual content selectivity in such electrodes.
Assuming that the ultra-fast ‘‘glow’’ does not inform about a specific content, it is tempting to speculate that a potential
(not exclusive) role of this global signal is to inform the non-relevant cortical areas about the occurrence of a high activation
event in content-selective cortical sites.
In this way, the low magnitude but rapid and wide-spread signaling may be relevant to a fundamental conundrum of all
strictly localist models, which concerns the mechanism that endows different cortical assemblies with their specific content.
To put the question simply: what is the essential change that occurs when neuronal activity moves from one local assembly
of neurons to a neighboring and rather similar looking one, that leads to the concomitant change in subjective perceptual
content (e.g. changing from a percept of a hand to a percept of a face)?
One obvious possible mechanism could be local differences in neuronal types or connectivity within each isolated assembly that somehow endows it with its subjective ‘‘quality’’ (e.g. face or hand). An alternative possibility (Tononi & Koch, 2008)
is that what underlies the different perceptual contents is the location of each neuronal assembly within a larger, global cortical framework or state space. Under this, admittedly highly speculative formulation, the global glow is a dynamic means for
this cortical framework to be rapidly established. One strong prediction of this conjecture is that such ultra-fast glow should
be a general phenomenon associated with the emergence of each and every new conscious event.
It is important to emphasize here a subtle point that is often ignored when considering ‘‘visual responses’’. The presence
of the ultra-fast global activation spread was revealed when visual image presentation was contrasted with a fixation baseline. It should be noted that in both the image presentation and fixation-baseline conditions the patients experienced a welldefined visual percept. The difference between these conditions was not in the presence of a subjective visual experience, but
rather, presumably, in the informational content of the subjective state – i.e., during image presentation the patients
observed meaningful images rich with detail, while during the baseline period the patients merely perceived a fixation cross.
Thus, if indeed a putative role of the global ‘‘glow’’ is to inform the cortex about the existence of a visual event, this signal is
clearly modulated by information-level aspects of the presented stimuli (such as the abrupt appearance, novelty, meaning,
arousing levels, semantic values, etc.) rather than the mere fact that they elicited a conscious visual percept (since such a
percept was also present during the fixation baseline).
4.4. Relevance to conscious perception
It is important to clarify that the present results do not prove that activation in high order visual areas is sufficient for
perceptual awareness to occur. A powerful method to investigate this point is based on modulating the content of visual
222
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
percepts while keeping the physical stimulus unchanged. This approach was not applied in the present study. However, previous work by our group and by others, using methods such as backward masking and binocular rivalry, demonstrated a
clear association between high activity in category-selective visual areas and perceptual awareness (e.g. Del Cul et al.,
2007; Fisch et al., 2009; Frässle et al., 2014).
Yet this contrast-based approach suffers from its own limitations. A central one is the fact that, in most cases, the perceptual changes involve rather selective changes in content. For example, in backward masking, the contrast between the
‘‘seeing’’ condition and ‘‘not seeing’’ condition is truly a contrast between perceiving a brief target (e.g. a face image) followed
by a mask – a highly visible and detailed visual stimulus vs. perceiving just the mask. Such contrasts can be detected either in
highly content-selective cortical sites, which as our current study demonstrates are confined mainly to high order visual
areas, or by subtracting the two conditions’ responses – which is methodologically problematic due to signal non-linearity
and the short temporal separation between target and mask.
Whether the content-invariant ‘‘glow’ depends on visual information that is consciously perceived or merely physically
presented, remains an interesting question to be explored. Given the difficulty in designing paradigms that will generate
drastic manipulations of visual informational content in the absence of a corresponding change in the physical stimulus,
answering it will be rather challenging.
An important point illustrated in Fig. 1 and discussed extensively above, is that when considering the issue of necessity
(i.e. whether high activity in fronto-parietal cortex is necessary for conscious visual perception), important insights can also
be gained by modulating post-visual aspects (e.g. motor report) while maintaining the visual percepts largely constant. Note
that, with this question specifically in mind, all of the alternatives proposed in Fig. 1 are testable without the use of perceptual manipulations.
4.5. Comparing the dynamics of BHF and ERP responses
The relatively short latency onset of visual responses across the cortex may appear to be at odds with the longer (300 ms)
latency that has been consistently reported to be associated with visual perception in scalp EEG studies (Dehaene &
Changeux, 2011; Del Cul et al., 2007; Sergent et al., 2005). However, our results appear to reconcile this apparent dichotomy.
This can be readily appreciated by comparing Figs. 4 and 5, showing the dynamics of the BHF and ERP responses. Note that
while the BHF response onset is at short (<200 ms) latencies, the ERP activity in non-visual cortex electrodes tends to gradually increase until it reaches its full peak activity around 300 ms. It is quite likely that during scalp recordings the dominant
features that could be discerned are these response peaks that tend to occur at 300 ms (but see also Foxe & Simpson, 2002).
4.6. The measurement window
In interpreting the present results, it is important to emphasize that ECoG recordings are limited to a specific spatiotemporal ‘‘window’’ of the neuronal activity. Such recordings typically do not contact deep sulci, leaving a substantial
expanse of fronto-parietal cortex inaccessible.
In addition, sparsely distributed neurons, i.e. active neurons that are not clustered in large groups, may produce signals
that are too weak to be discernible in the ECoG signals which reflect averaged mass activity. Hence, caution should be exercised when comparing ECoG to BOLD-fMRI which integrate neuronal activity over larger spatial expanses and temporal durations and is more sensitive to sparse activity (Nir, Dinstein, Malach, & Heeger, 2008; Privman et al., 2007).
Thus, our results certainly do not rule out a number of additional possible routes by which fronto-parietal activation may
occur during conscious vision. However, they do suggest that perceptual awareness, as reflected in the ability of the patients
to categorize and remember the visual stimuli, can emerge even when the global signals that spread to fronto-parietal cortex
are of low magnitude, reflecting weak activations or sparse neuronal responsiveness. Such low-magnitude signals should be
distinguished from the high magnitude gamma ‘‘ignitions’’ found to be associated with the emergence of a visual percept in
high order visual areas (Fisch et al., 2009; Lachaux et al., 2005).
4.7. A hybrid model
In summary, our results appear to be incompatible with the strictly local or global alternatives suggested in Fig. 1. On the
one hand, they are indeed compatible with the local alternative which predicts content specific, high magnitude activation
confined to the visual cortex proper. On the other hand, they also reveal an additional global aspect of the neuronal response,
instantiated in ultra-fast spread of low magnitude, largely content-invariant signals into selected sites of fronto-parietal cortex. The precise functional role of this global ‘‘glow’’ remains to be elucidated.
Acknowledgments
This research was supported by the EU FP7 VERE, EU-Flagship HBP, ICORE program (ISF 51/11), Page and Otto Marx Jr.
Foundation for SB and ADM, the Helen and Martin Kimmel award to R.M., National Institute of Mental Health grant
MH093061 to E.Z and MH086385 to C.E.S. We are grateful to the six patients who participated in this study.
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
223
References
Baars, B. J. (1997). In the theater of consciousness: The workspace of the mind. Oxford University Press.
Baars, B. J. (2002). The conscious access hypothesis: Origins and recent evidence. Trends in Cognitive Sciences, 6(1), 47–52. http://dx.doi.org/10.1016/S13646613(00)01819-2.
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal
Statistical Society: Series B (Methodological), 57(1), 289–300.
Block, N. (2005). Two neural correlates of consciousness. Trends in Cognitive Sciences, 9(2), 46–52. http://dx.doi.org/10.1016/j.tics.2004.12.006.
Block, N. (2007). Consciousness, accessibility, and the mesh between psychology and neuroscience. Behavioral and Brain Sciences, 30(5), 481–498. http://
dx.doi.org/10.1017/S0140525X07002786.
Davidesco, I., Harel, M., Ramot, M., Kramer, U., Kipervasser, S., Andelman, F., et al (2013a). Spatial and object-based attention modulates broadband highfrequency responses across the human visual cortical hierarchy. The Journal of Neuroscience, 33(3), 1228–1240. http://dx.doi.org/10.1523/
JNEUROSCI.3181-12.2013.
Davidesco, I., Zion-Golumbic, E., Bickel, S., Harel, M., Groppe, D. M., Keller, C. J., et al (2013b). Exemplar selectivity reflects perceptual similarities in the
human fusiform cortex. Cerebral Cortex, 24(7), 1879–1893. http://dx.doi.org/10.1093/cercor/bht038.
Dehaene, S., & Changeux, J. P. (2011). Experimental and theoretical approaches to conscious processing. Neuron, 70(2), 200–227. http://dx.doi.org/10.1016/
j.neuron.2011.03.018.
Dehaene, S., Changeux, J., Naccache, L., Sackur, J., & Sergent, C. (2006). Conscious, preconscious, and subliminal processing: A testable taxonomy. Trends in
Cognitive Sciences, 10(5), 204–211. http://dx.doi.org/10.1016/j.tics.2006.03.007.
Del Cul, A., Baillet, S., & Dehaene, S. (2007). Brain dynamics underlying the nonlinear threshold for access to consciousness. PLoS Biology, 5(10), e260. http://
dx.doi.org/10.1371/journal.pbio.0050260.
Delorme, A., & Makeig, S. (2004). EEGLAB: An open source toolbox for analysis of single-trial EEG dynamics including independent component analysis.
Journal of Neuroscience Methods, 134(1), 9–21. http://dx.doi.org/10.1016/j.jneumeth.2003.10.009.
Fedorenko, E., Duncan, J., & Kanwisher, N. (2013). Broad domain generality in focal regions of frontal and parietal cortex. Proceedings of the National Academy
of Sciences, 110(41), 16616–16621. http://dx.doi.org/10.1073/pnas.1315235110.
Fisch, L., Privman, E., Ramot, M., Harel, M., Nir, Y., Kipervasser, S., et al (2009). Neural ‘‘Ignition’’: Enhanced activation linked to perceptual awareness in
human ventral stream visual cortex. Neuron, 64(4), 562–574. http://dx.doi.org/10.1016/j.neuron.2009.11.001.
Foxe, J. J., & Simpson, G. V. (2002). Flow of activation from V1 to frontal cortex in humans. Experimental Brain Research, 142(1), 139–150. http://dx.doi.org/
10.1007/s00221-001-0906-7.
Frässle, S., Sommer, J., Jansen, A., Naber, M., & Einhäuser, W. (2014). Binocular rivalry: Frontal activity relates to introspection and action but not to
perception. The Journal of neuroscience, 34(5), 1738–1747. http://dx.doi.org/10.1523/JNEUROSCI.4403-13.2014.
Gaillard, R., Dehaene, S., Adam, C., Clémenceau, S., Hasboun, D., Baulac, M., et al (2009). Converging intracranial markers of conscious access. PLoS Biology,
7(3), e1000061. http://dx.doi.org/10.1371/journal.pbio.1000061.
Grill-Spector, K., & Malach, R. (2001). FMR-adaptation: A tool for studying the functional properties of human cortical neurons. Acta Psychologica, 107(1),
293–321. http://dx.doi.org/10.1016/S0001-6918(01)00019-1.
Henrie, J. A., & Shapley, R. (2005). LFP power spectra in V1 cortex: The graded effect of stimulus contrast. Journal of Neurophysiology, 94(1), 479–490. http://
dx.doi.org/10.1152/jn.00919.2004.
Koivisto, M., & Revonsuo, A. (2010). Event-related brain potential correlates of visual awareness. Neuroscience and Biobehavioral Reviews, 34(6), 922–934.
http://dx.doi.org/10.1016/j.neubiorev.2009.12.002.
Kouider, S., Stahlhut, C., Gelskov, S. V., Barbosa, L. S., Dutat, M., de Gardelle, V., et al (2013). A neural marker of perceptual consciousness in infants. Science,
340(6130), 376–380. http://dx.doi.org/10.1126/science.1232509.
Lachaux, J. P., George, N., Tallon-Baudry, C., Martinerie, J., Hugueville, L., Minotti, L., et al (2005). The many faces of the gamma band response to complex
visual stimuli. Neuroimage, 25(2), 491–501. http://dx.doi.org/10.1016/j.neuroimage.2004.11.052.
Lamme, V. A. F., & Roelfsema, P. R. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in Neurosciences, 23(11),
571–579. http://dx.doi.org/10.1016/S0166-2236(00)01657-X.
Lamy, D., Salti, M., & Bar-Haim, Y. (2009). Neural correlates of subjective awareness and unconscious processing: An ERP study. Journal of Cognitive
Neuroscience, 21(7), 1435–1446. http://dx.doi.org/10.1162/jocn.2009.21064.
Malach, R. (2007). The measurement problem in consciousness research. Behavioral and Brain Sciences, 30(5), 516–517. http://dx.doi.org/10.1017/
S0140525X0700297X.
Mehta, A. D., Ulbert, I., & Schroeder, C. E. (2000a). Intermodal selective attention in monkeys. I: Distribution and timing of effects across visual areas. Cerebral
Cortex, 10(4), 343–358. http://dx.doi.org/10.1093/cercor/10.4.343.
Mehta, A. D., Ulbert, I., & Schroeder, C. E. (2000b). Intermodal selective attention in monkeys. II: Physiological mechanisms of modulation. Cerebral Cortex,
10(4), 359–370. http://dx.doi.org/10.1093/cercor/10.4.359.
Meshulam, M., Ramot, M., Harel, M., Kipervasser, S., Andelman, F., Neufeld, M. Y., et al (2013). Selectivity of audiovisual ECoG responses revealed under
naturalistic stimuli in the human cortex. Journal of Neurophysiology, 109(9), 2272–2281. http://dx.doi.org/10.1152/jn.00474.2012.
Mukamel, R., Gelbard, H., Arieli, A., Hasson, U., Fried, I., & Malach, R. (2005). Coupling between neuronal firing, field potentials, and FMRI in human auditory
cortex. Science, 309(5736), 951–954. http://dx.doi.org/10.1126/science.1110913.
Nir, Y., Dinstein, I., Malach, R., & Heeger, D. J. (2008). BOLD and spiking activity. Nature Neuroscience, 11(5), 523–524. http://dx.doi.org/10.1038/nn0508-523.
Nir, Y., Fisch, L., Mukamel, R., Gelbard-Sagiv, H., Arieli, A., Fried, I., et al (2007). Coupling between neuronal firing rate, gamma LFP, and BOLD fMRI is related
to interneuronal correlations. Current Biology, 17(15), 1275–1285. http://dx.doi.org/10.1016/j.cub.2007.06.066.
Pins, D., & Ffytche, D. (2003). The neural correlates of conscious vision. Cerebral Cortex, 13(5), 461–474. http://dx.doi.org/10.1093/cercor/13.5.461.
Pitts, M. A., Metzler, S., & Hillyard, S. A. (2014). Isolating neural correlates of conscious perception from neural correlates of reporting one’s perception.
Frontiers in Psychology, 5. http://dx.doi.org/10.3389/fpsyg.2014.01078.
Privman, E., Nir, Y., Kramer, U., Kipervasser, S., Andelman, F., Neufeld, M. Y., et al (2007). Enhanced category tuning revealed by intracranial
electroencephalograms in high-order human visual areas. The Journal of neuroscience, 27(23), 6234–6242. http://dx.doi.org/10.1523/JNEUROSCI.462706.2007.
Ramot, M., Fisch, L., Harel, M., Kipervasser, S., Andelman, F., Neufeld, M. Y., et al (2012). A widely distributed spectral signature of task-negative
electrocorticography responses revealed during a visuomotor task in the human cortex. The Journal of neuroscience, 32(31), 10458–10469. http://
dx.doi.org/10.1523/JNEUROSCI.0877-12.2012.
Ray, S., & Maunsell, J. H. R. (2011). Different origins of gamma rhythm and high-gamma activity in macaque visual cortex. PLoS Biology, 9(4), e1000610.
http://dx.doi.org/10.1371/journal.pbio.1000610.
Schroeder, C., Mehta, A., & Givre, S. (1998). A spatiotemporal profile of visual system activation revealed by current source density analysis in the awake
macaque. Cerebral Cortex, 8(7), 575–592. http://dx.doi.org/10.1093/cercor/8.7.575.
Sergent, C., Baillet, S., & Dehaene, S. (2005). Timing of the brain events underlying access to consciousness during the attentional blink. Nature Neuroscience,
8(10), 1391–1400. http://dx.doi.org/10.1038/nn1549.
Sherman, S. M., & Guillery, R. (1996). Functional organization of thalamocortical relays. Journal of Neurophysiology, 76(3), 1367–1395.
Talairach, J., & Tournoux, P. (1988). Co-planar stereotaxic atlas of the human brain: 3-dimensional proportional system: An approach to cerebral imaging: Thieme.
Tallon-Baudry, C., & Bertrand, O. (1999). Oscillatory gamma activity in humans and its role in object representation. Trends in Cognitive Sciences, 3(4),
151–162. http://dx.doi.org/10.1016/S1364-6613(99)01299-1.
224
N. Noy et al. / Consciousness and Cognition 35 (2015) 206–224
Tononi, G., & Koch, C. (2008). The neural correlates of consciousness. Annals of the New York Academy of Sciences, 1124(1), 239–261. http://dx.doi.org/
10.1196/annals.1440.004.
Vidal, J. R., Ossandón, T., Jerbi, K., Dalal, S. S., Minotti, L., Ryvlin, P., et al (2010). Category-specific visual responses: An intracranial study comparing gamma,
beta, alpha, and ERP response selectivity. Frontiers in Human Neuroscience, 4. http://dx.doi.org/10.3389/fnhum.2010.00195.
Zeki, S. (2001). Localization and globalization in conscious vision. Annual Review of Neuroscience, 24(1), 57–86. http://dx.doi.org/10.1146/
annurev.neuro.24.1.57.
Zeki, S. (2003). The disunity of consciousness. Trends in Cognitive Sciences, 7(5), 214–218. http://dx.doi.org/10.1016/S1364-6613(03)00081-0.
Zeki, S., & Bartels, A. (1999). Toward a theory of visual consciousness. Consciousness and Cognition, 8(2), 225–259. http://dx.doi.org/10.1006/ccog.1999.0390.