Available online at www.sciencedirect.

com

ScienceDirect

Toward a computational theory of conscious processing

Stanislas Dehaene1,2,3,4, Lucie Charles2,3,4, Jean-Rémi King2,3,4 and
Sébastien Marti2,3,4
The study of the mechanisms of conscious processing has others remain non-conscious? Considerable empirical
become a productive area of cognitive neuroscience. Here we and theoretical progress has been made lately in answer-
review some of the recent behavioral and neuroscience data, ing this deceivingly simple question. This research
with the specific goal of constraining present and future gained leverage when it was recognized that visual illu-
theories of the computations underlying conscious processing. sions [1–3] and a great variety of other normal and
Experimental findings imply that most of the brain’s pathophysiological conditions such sleep, anesthesia,
computations can be performed in a non-conscious mode, but blindsight or hemineglect provided empirical windows
that conscious perception is characterized by an amplification, into this phenomenon, by providing minimal contrasts
global propagation and integration of brain signals. A between conscious and non-conscious brain states [4].
comparison of these data with major theoretical proposals Here we review the recent advances made possible by
suggests that firstly, conscious access must be carefully this contrastive approach. We specifically focus on how
distinguished from selective attention; secondly, conscious these findings inform present-day theories of conscious
perception may be likened to a non-linear decision that ‘ignites’ processing. At present, there is no accepted compu-
a network of distributed areas; thirdly, information which is tational theory of this function. Our hope is that the
selected for conscious perception gains access to additional present review may point to the key ingredients that will
computations, including temporary maintenance, global lead to one.
sharing, and flexible routing; and finally, measures of the
complexity, long-distance correlation and integration of brain Defining the terms
signals provide reliable indices of conscious processing, It is useful to start by separating the diversity of concepts
clinically relevant to patients recovering from coma. that the everyday term of ‘consciousness’ can refer to.
Addresses The content of consciousness refers to the specific infor-
1
Collège de France, F-75005 Paris, France mation that I am aware of at a given moment. For
2
Cognitive Neuroimaging Unit, Institut National de la Santé et de la
Recherche Médicale, U992, F-91191 Gif/Yvette, France
instance, I am currently aware of reading these words,
3
NeuroSpin Center, Institute of BioImaging Commissariat à l’Energie but not of the music playing in the background (until I
Atomique, F-91191 Gif/Yvette, France attend to it). Conscious access is the process by which a
4
Université Paris 11, Orsay, France piece of information becomes a conscious content. Con-
scious processing refers to the various operations that can be
Corresponding author: Dehaene, Stanislas
([email protected]) applied to a conscious content (as when multiplying two
numbers mentally). Conscious report is the process by
which a conscious content can be described, verbally or
Current Opinion in Neurobiology 2014, 25:76–84 by various gestures. Such reportability remains the main
This review comes from a themed issue on Theoretical and criterion for whether a piece of information is or is not
computational neuroscience conscious: by hypothesis, I can report something if and
Edited by Adrienne Fairhall and Haim Sompolinsky only if I am aware of it.

A great variety of representations can be consciously

accessed, including perceptual states, abstract knowl-
0959-4388/$ – see front matter, # 2013 Elsevier Ltd. All rights
reserved.
edge, memories, plans, and other internal states (e.g.
feelings, confidence, and errors). Self-consciousness is a
http://dx.doi.org/10.1016/j.conb.2013.12.005
particular instance of conscious access where the con-
scious ‘spotlight’ is oriented toward internal states.

Introduction The state of consciousness, associated with fluctuations in

wakefulness or vigilance, finally, refers to the brain’s very
ability to entertain a stream of conscious contents. During
Consciousness is the only real thing in the world and the
normal wakefulness, any information may be consciously
greatest mystery of all
accessed, but this ability is continuously modulated
Vladimir Nabokov, Bend Sinister (1947)
according to the level of vigilance, and ultimately
vanishes during coma, vegetative state, anesthesia or
What brain mechanisms underlie our capacity to become deep sleep. Although this review concentrates primarily
aware of a specific piece of information, while many on the mechanisms of conscious access and conscious

Current Opinion in Neurobiology 2014, 25:76–84 www.sciencedirect.com

 Theorizing conscious processing Dehaene et al. 77

processing, in a final section, we consider how what has signatures that occur on distinct trials and at different
been learned about conscious access in normal subjects times [28–30]. For instance, by orthogonally manipulating
generalizes to the detection of the state of consciousness visibility and attention (using masked images presented
in brain-lesioned patients. at the threshold for conscious perception such that half
were visible and half were invisible, and preceding them
The boundaries of non-conscious processing by valid or invalid attentional cues), Wyart and colleagues
To clarify the nature of conscious processing, a first step [29] found a double dissociation: attention, but not visi-
consists in delineating what it is not. Using masking [5], bility, modulated early occipital activity, while visibility,
crowding [6], inattention [7] or binocular rivalry [8], but not attention, modulated later temporal and parieto-
images can be presented under conditions such that they frontal activity. Under some circumstances, greater spatial
remain strictly invisible. Behavioral priming and brain attention may even lead to a reduced visibility [31].
imaging can then reveal how deep these stimuli are These findings refute theories that conflate attention and
processed. Studies of non-conscious processing play an consciousness. William James’ classical definition of
instrumental role in refuting specific theories of con- attention (‘the taking possession by the mind, in clear
sciousness. The logic is simple: if a cognitive computation and vivid form, of one out of what seem several simul-
or neural marker, proposed by some theory to be uniquely taneously possible objects or trains of thought’) mixes up
associated with conscious processing, can be observed conscious access proper (‘taking possession of the mind’)
under demonstrably non-conscious conditions, then that with selection (‘one out of several’) which can be fully
theory is severely undermined. non-conscious. Selective attention may facilitate con-
scious access, even when the attentional cue comes long
Twenty years of research indicates that subliminal pro- after the stimulus is gone [32], but it operates largely
cessing can be quite deep. Many cortical areas can be non-consciously.
activated by an unseen stimulus, including areas of the
visual ventral [9] and dorsal pathways [10]. The brain non- Recent findings also invalidate the idea that the central
consciously recognizes the abstract identity of pictures, executive, which controls our strategies and inhibits
words and faces [9,11,12], the quantity attached to a unwanted behaviors, always operates consciously. A
number symbol [10,13], the fact that two words are series of experiments with the go/no-go paradigm indicate
related or synonymous [6,14,15], the emotional meaning that an unseen visual cue can trigger inhibitory control
of a word [16,17], or the reward value of a coin or an circuits in the pre-supplementary motor area and anterior
arbitrary symbol [18,19,20]. insula [33,34,35,36]. Error detection [37,38] and task
switching [39,40], which are typical executive functions,
In recent years, the frontiers of non-conscious processing can be triggered non-consciously. Even the maintenance
have been pushed further. For instance, in chess experts, of a stimulus in working memory may remain above the
a brief non-conscious flash of a chessboard suffices to chance level for subliminal stimuli [41] — although this
determine whether the king is in check [21]. Within the recent finding will need to be reconciled with the more
language domain, the grammatical fit of a masked word frequent observation that subliminal priming drops to
with the preceding sentence can be determined non- chance level after a second or less [42–44].
consciously [22]. Transitive inferences can also be
deployed non-consciously: after non-conscious exposure Overall, these findings support the view that virtually any
to arbitrary word pairs such as ‘winter-red’ and ‘red- cerebral processor may operate in a non-conscious mode.
computer’, word association effects generalize to non- They challenge theories that associate conscious proces-
adjacent pairs (‘winter-computer’), a transitive link sing with a specific cognitive processor. For instance, the
mediated by the hippocampus [23]. As another example hypothesis that conscious perception coincides with the
of high-level computation, the approximate average of ability to deploy higher-order thoughts or metacognition
four masked numbers can be extracted non-consciously (the brain’s ability to represent its own knowledge states)
[13]. There is even a suggestion that multi-step oper- [45] does not bode well with evidence that self-monitor-
ations such ‘9 5 + 2’ may be mediated non-consciously ing, error detection and confidence assignment partially
[24], although this conclusion will require better control operate non-consciously [38,46,47].
over the stimuli and the degree of non-consciousness.
Findings from subliminal research also eliminate some
All in all, these findings refute the idea that non-conscious physiological theories of conscious processing. It is now
processing stops at an early perceptual level: meaning and clear that early changes in gamma band power (>30 Hz),
value can clearly be assigned non-consciously. There is once postulated as a marker of consciousness, can be
also considerable evidence that attention can be deployed evoked by a non-conscious stimulus [48,49] and do
and enhance processing even if its target remains non- not faithfully track variations in subjective reports [50].
conscious [25–27]. At the brain level, attending to a Similarly, the views that recurrent interactions [51,52] and
stimulus and becoming conscious of it have distinct information integration [53,54] are necessary and

www.sciencedirect.com Current Opinion in Neurobiology 2014, 25:76–84

78 Theoretical and computational neuroscience

sufficient markers of conscious processing, although not connections impose slow accumulation dynamics and
directly refuted, are made implausible by empirical find- multi-stable ‘all-or-none’ behavior, whereby the incom-
ings of non-conscious interactions between frontal and ing evidence either quickly dies out (corresponding to
occipital regions [55], non-conscious integration of subliminal processing) or is accumulated and amplified
unseen visual contours [56], unseen objects in an unseen non-linearly into a full-blown state of high-level activity.
complex visual scene [57], or unseen words in the seman- This global ‘ignition’ has been proposed as a marker of
tic or syntactic context of other words [14,22]. These conscious perception [70]. Indeed, empirically, when
operations are slow (220–260 ms for contour integration, stimulus strength is varied, the early stages of non-con-
400 ms for semantic integration), clearly involve integ- scious processing typically show a linear variation in
ration of multiple sources, and are unlikely to occur in a activation, whereas conscious access is often character-
purely feedforward manner without recurrent inter- ized by a late non-linear amplification of activation which
actions, and yet they occur non-consciously. Similarly, invades a distributed set of parietal, prefrontal and cin-
serial accumulation of evidence can occur without aware- gulate areas [58,60,61, for extensive review, see 70–
ness [58,59]. 72,73,74]. In behavior, perceptual processing is continu-
ous for subliminal stimuli, but becomes categorical when
Conscious access as an accumulation of the stimulus is seen [75,76]. In EEG, MEG, and intra-
evidence leading to an all-or-none ignition cranial recordings, conscious stimuli, compared to
What, if anything, remains unique to conscious proces- matched non-conscious ones, induce a late (300 ms)
sing? Although many cognitive operations can be partially and sudden increase in slow event-related potentials
launched non-consciously, these operations rarely if ever (inducing a P3 wave on the scalp), gamma power and
run to completion in the absence of consciousness. A long-range beta and gamma synchrony [48,49,77].
subliminal stimulus may induce above-chance perform- Specific components such as the error-related negativity
ance, behavioral priming, and a small amount of brain evoked after an erroneous motor response also follow this
activity in narrowly defined brain networks, but these ‘all-or-none’ non-linear pattern [46,78].
measures often increase dramatically as soon as the sub-
ject reports seeing the stimulus, especially in high-level A direct relation between evidence accumulation and
areas [46,60,61]. Accumulation of evidence has been conscious visibility was demonstrated in a recent MEG
demonstrated with non-conscious stimuli [59], but only experiment with gratings presented at threshold [79].
conscious stimuli cross the threshold beyond which an The subjective reports of seeing or not-seeing could be
overt strategy can be flexibly deployed [58]. predicted on a single trial basis as a sum of gamma power
present before the presentation of the stimulus ( 300 to
Such findings vindicate the pre-theoretical idea that 100 ms) and long after it (+250 to +450 ms). Thus,
consciousness possesses a threshold that separates sublim- whether a stimulus is detected seemed to be determined
inal and supraliminal stimuli (limen is the Latin word for by an accumulation of pre-stimulus bias (‘prior’) and
threshold). Several theorists propose that conscious per- stimulus-evoked activation (‘evidence’) [see also 80].
ception occurs when the stimulus allows the accumu-
lation of sufficient sensory evidence to reach a threshold, Late ignition seems to provide a robust signature of
at which point the brain ‘decides’ whether it has seen conscious access. The contrast between an early linear
anything, and what it is [62,63]. The mechanisms of variation in brain activity and a very late non-linear
conscious access would then be comparable to those of ignition has even been observed in 5, 12 and 15-
other decisions, involving an accumulation toward a month-old infants [81], leading to the tentative sugges-
threshold — with the difference that conscious percep- tion that infants too enjoy a conscious perception of visual
tion would correspond to a global high-level ‘decision to stimuli, albeit at a much slower pace.
engage’ many of the brain’s internal resources, not just a
single effector [63]. The mathematical frameworks of It remains debated, however, whether ignition is a unified
signal detection theory and Bayesian decision making process or whether it can be decomposed into a series of
have been used to model subjective reports of seeing stages that correspond to pre-conscious, conscious and
in normal subjects and blindsight patients [64,65]. Neural post-conscious processes [82]. The P3 wave may partly
network models have also been proposed for how high- reflect processes that unfold after conscious access, such
order cortices might accumulate metacognitive evidence as executive attention, working memory updating, or the
about the state of other cortices, rather than about the preparation of a behavioral report. When these processes
external world, leading to a confident feeling of seeing are eliminated by making the stimulus irrelevant to the
[66]. current task, its conscious perception may correlate solely
with a transient posterior negativity of moderate size,
Recurrent thalamo-cortical networks provide a simple peaking around 300 ms [56,83], although other studies
and generic implementation of elementary stimulus continue to observe a large and long-lasting effect
categorization processes [67–69]. Recurrent NMDA [84,85].

Current Opinion in Neurobiology 2014, 25:76–84 www.sciencedirect.com

 Theorizing conscious processing Dehaene et al. 79

Conscious processing as global information primes can have cumulative non-conscious effects on a
sharing behavioral decision, but only conscious primes allow for
Global Neuronal Workspace (GNW) theory [2,86,87] the development of subsequent serial strategies
proposes that conscious access stems from a cognitive [58,92,93].
architecture with an evolved function: the flexible sharing
of information throughout the cortex [4]. While non- The brain’s routing system is capacity-limited, and this
conscious stimuli are processed in parallel by specialized feature may explain the frequent failure of conscious
cortical processors, conscious perception would be perception in a dual-task setting. Conscious processing
needed in order to flexibly route a selected stimulus of a first target T1 causes a bottleneck on the routing of a
through a series of non-routine information processing subsequent target T2, either by dramatically postponing
stages. Global information sharing and routing would rely its processing (a phenomenon known as the ‘psychologi-
on a set of interconnected high-level cortical regions cal refractory period’, PRP) or by preventing its conscious
forming a ‘global workspace’ and involving primarily perception altogether (‘attentional blink’, AB). Recent
the dorsolateral prefrontal cortex, but also additional hubs evidence confirms that PRP and AB are tightly related
in inferior parietal cortex, mid-temporal cortex, and pre- phenomena that may occur within the same experiment
cuneus, and now described as forming a ‘rich club’ net- [94]. Like AB, PRP causes a loss of conscious perception:
work [88,89]. the second target T2 is not only delayed, but also tempor-
arily unperceived, such that its subjective onset is dis-
Behavioral research supports this idea in various ways. A placed to the moment when T1 processing finishes [95].
subliminal prime often facilitates performance in a single The minimal condition for creating these effects is that
task, but this non-conscious performance drops to chance T1 is consciously perceived [96,97]. These effects have
level when the task requires a series of novel operations been related to a global parietal and prefrontal network
that involve ‘piping’ the output of one process to the [94,98], and have been partially captured in simulations of
input of another [90,91]. Likewise, a series of subliminal spiking neurons [69,99] (Figure 1).

Figure 1

(a) 70 (e)

Evaluative st
Sensory 1
Systems
(VALUE) 0
60

Sensory 2nd

0
70
Long-Term Attentional Sensory 3rd
Memory Systems
(PAST) (FOCUSING)
0
Global (b) (f)
40
Workspace
Router

0
(c) 80 (g)
Task-setting

0
Motor (d) (h)
150
spikes / sec

Perceptual systems
(FUTURE) Response
systems
(PRESENT)
0
0 200 400 600 0 200 400 600
Time (ms)
Current Opinion in Neurobiology

Flexible information routing and conscious processing in large-scale models of the cortex. (left) Original depiction of the Dehaene-Changeux model of
a Global Neuronal Workspace [GNW; Ref. 87]. The GNW model proposes that what we subjectively experience as a conscious state is the global
availability of the corresponding information. Conscious access would occur when a piece of information enters a distributed network of cortical areas
tightly interconnected by long-distance axons, the GNW, which allows its flexible broadcasting to any of the brain’s many specialized processors.
(right) Spiking-neuron simulation of a flexible routing system [Ref. 99]. While a first stimulus is processed and routed to an arbitrary response (left
column, top to bottom), a second stimulus (right column) is also processed perceptually but is then blocked at the routing stage. This model captures
in great detail two neuropsychological phenomena, the psychological refractory period and the attentional blink.

www.sciencedirect.com Current Opinion in Neurobiology 2014, 25:76–84

80 Theoretical and computational neuroscience

In order to be globally shared, conscious information anesthetics are captured by a simple neuronal network
should be represented by a stable and reproducible model [109,110].
brain-scale assembly for a minimal duration. This stability
criterion was explicitly tested in an fMRI study where Most importantly, integration and long-distance cortical
brain activity patterns were more reproducible across communications provide signatures of residual conscious-
trials for perceived than for unperceived stimuli [100]. ness that are clinically applicable to patients recovering
Electro-encephalography and magneto-encephalography from coma. From behavior alone, the presence of con-
confirm that conscious processing causes sustained brain sciousness may be quite difficult to detect, and functional
activity, often extending for several hundreds of milli- MRI has revealed that a few patients in apparent vege-
seconds [29,84,85,101,102]. In intracranial recordings, tative state may, in fact, be fully conscious and ‘locked-in’
conscious stimuli, but not non-conscious ones, trigger a [111]. An exciting study indicates that the complexity of
sustained activation and the formation of a metastable EEG waves evoked by a single TMS pulse to the cortex
state of long-distance phase synchrony in the beta band provides a strictly quantitative measure of the state of
[48]. Nevertheless, a debate remains, as some data consciousness, with a bimodal distribution separating the
suggest that local synchrony and reverberation may suf- awake state from sleep, anesthesia, coma or vegetative
fice for conscious perception [49,103], while other state [112]. Similarly, an EEG measure of the amount of
experiments indicate that long-distance synchrony be- information shared by distant cortical sites provides a
tween prefrontal and occipital cortex may exist even highly sensitive discrimination of patients in vegetative
under non-conscious conditions [55]. versus minimally conscious states, regardless of etiology
and time elapsed since injury [113]. Both observations
Consciousness as integrated information suggest that global cortical communication provides an
According to Information Integration Theory (IIT) excellent index of conscious processing, and are in agree-
[53,54], global synchrony and re-entry may be needed, ment with both GNW and IIT theories.
not just to globally share or broadcast a conscious
stimulus, but, more essentially, to create an integrated Conclusion
representation of its various features. A precise math- Consciousness research has truly come of age. Empiri-
ematical formula (F is proposed to quantify the amount of cally, several candidate markers of conscious processing
integration of a system composed of multiple parts). High are now available. Theoretically, we reviewed three
levels of F would be indicative of a conscious device specific theoretical proposals that tentatively relate con-
(whether biological or artificial). Any system would pos- scious processing, respectively, to global ignition, long-
sess a small amount of F and therefore some degree of distance broadcasting, and information integration.
consciousness (panpsychism). This formal framework is These ideas are not necessarily incompatible. On the
however limited in its ability to make specific behavioral contrary, considerable convergence exists to suggest that
and biological predictions. Indeed, F is impossible to firstly, conscious access triggers an all-or-none change in
compute in practice (only approximations exist [104]). the state of distributed cortical networks; secondly, con-
Furthermore, this theory does not offer any neurophy- scious processing corresponds to a massive cortico-cortical
siological mechanisms for why conscious perception fol- exchange of information, allowing flexible routing and
lows a non-linear profile or why highly integrative therefore the slow serial performance of novel and arbi-
semantic processes can be triggered non-consciously, as trary tasks; and finally, the state of consciousness, that is
reviewed above. the brain’s very ability to host a ceaseless stream of such
all-or-none conscious episodes, rests upon the integrity of
A more modest proposal is that F and related quantities long-distance cortico-cortical exchanges, which can be
provide one of many possible signatures of the state of continuously modulated by lesions or anesthetics and is
consciousness [104,105], simply because they reflect the reflected by electrophysiological indices of brain-wide
brain’s capacity to broadcast information in the global information sharing.
neuronal workspace, and therefore to entertain a cease-
less stream of episodes of conscious access and conscious Future research should investigate whether the proposed
processing [70]. Experimentally, mathematical measures markers of conscious processing are generic and valid in
of the complexity and global integration of brain signals all conditions, or whether some are more diagnostic than
do provide solid markers of the state of consciousness, others. Above all, more detailed computational theories,
particularly when contrasting wakefulness with sleep or framed as large-scale simulations of spiking neurons, will
anesthesia. Intracranial recordings in humans undergoing be needed to understand the conditions of their emer-
propofol anesthesia indicate a dramatic and sudden frag- gence in experimental recordings.
mentation of neural activity, which remains locally orga-
nized but globally disintegrated [106, see also 107], Acknowledgements
possibly because prefrontal cortices are invaded by an We gratefully acknowledge extensive discussions with Jean-Pierre
alpha-like rhythm [108]. Some of these effects of Changeux, Lionel Naccache, and Aaron Schurger. This work was supported

Current Opinion in Neurobiology 2014, 25:76–84 www.sciencedirect.com

 Theorizing conscious processing Dehaene et al. 81

by a Direction Générale de l’Armement grant (to JRK), an European 17. Naccache L, Gaillard R, Adam C, Hasboun D, Clémenceau S,
Research Council grant ‘‘NeuroConsc’’ (to SD), and by the European Baulac M, Dehaene S, Cohen L: A direct intracranial record of
Human Brain Project. emotions evoked by subliminal words. Proc Natl Acad Sci U S A
2005, 102:7713-7717.
References and recommended reading 18. Schmidt L, Palminteri S, Lafargue G, Pessiglione M: Splitting
Papers of particular interest, published within the period of review, motivation: unilateral effects of subliminal incentives. Psychol
have been highlighted as: Sci 2010, 21:977-983.

 of special interest 19. Pessiglione M, Petrovic P, Daunizeau J, Palminteri S, Dolan RJ,

 of outstanding interest Frith CD: Subliminal instrumental conditioning demonstrated
in the human brain. Neuron 2008, 59:561-567.
1. Crick F, Koch C: Some reflections on visual awareness. Cold 20. Pessiglione M, Schmidt L, Draganski B, Kalisch R, Lau H,
Spring Harb Symp Quant Biol 1990, 55:953-962.  Dolan RJ, Frith CD: How the brain translates money into force: a
neuroimaging study of subliminal motivation. Science 2007,
2. Dehaene S, Naccache L: Towards a cognitive neuroscience of 316:904-906.
consciousness: basic evidence and a workspace framework. Participants were presented with masked images of coins serving as
Cognition 2001, 79:1-37. incentives while they performed a hand-grip force task. Recordings of
skin conductance, hand-grip force and brain activity indicated that
3. Kim CY, Blake R: Psychophysical magic: rendering the visible
subliminal incentives modulated the brain’s reward and motivation cir-
‘invisible’. Trends Cogn Sci 2005, 9(8):381-388.
cuits, as well as the participants’ behavioral responses.
4. Baars BJ: A Cognitive Theory of Consciousness. Cambridge, MA:
21. Kiesel A, Kunde W, Pohl C, Berner MP, Hoffmann J: Playing chess
Cambridge University Press; 1989, .
unconsciously. J Exp Psychol Learn Mem Cogn 2009,
5. Kouider S, Dehaene S: Levels of processing during non- 35:292-298.
conscious perception: a critical review of visual
22. Batterink L, Neville HJ: The human brain processes syntax in the
masking. Philos Trans R Soc Lond B: Biol Sci 2007,
362:857-875.  absence of conscious awareness. J Neurosci 2013, 33:8528-
8533.
6. Yeh SL, He S, Cavanagh P: Semantic priming from crowded In this study, syntactic violations in a stream of words were made
words. Psychol Sci 2012, 23:608-616. undetectable by a brief auditory distraction. The undetected violations
stilll evoked a early left anterior negativity, indicating that syntactic
7. Chun MM, Marois R: The dark side of visual attention. Curr Opin processing can proceed without awareness. Only detected violations
Neurobiol 2002, 12:184-189. evoked a later P600 response, confirming that late positive event-related
potentials are a frequent marker of conscious processing.
8. Pitts MA, Britz J: Insights from intermittent binocular rivalry and
EEG. Front Hum Neurosci 2011, 5:107. 23. Reber TP, Luechinger R, Boesiger P, Henke K: Unconscious
relational inference recruits the hippocampus. J Neurosci
9. Dehaene S, Naccache L, Cohen L, Le Bihan D, Mangin JF, 2012, 32:6138-6148.
Poline JB, Rivière D: Cerebral mechanisms of word masking
and unconscious repetition priming. Nat Neurosci 2001, 24. Sklar AY, Levy N, Goldstein A, Mandel R, Maril A, Hassin RR:
4:752-758. Reading and doing arithmetic nonconsciously. Proc Natl Acad
Sci U S A 2012, 109:19614-19619.
10. Naccache L, Dehaene S: The priming method: imaging
unconscious repetition priming reveals an abstract 25. Kentridge RW, Nijboer TC, Heywood CA: Attended but unse
representation of number in the parietal lobes. Cereb Cortex visual attention is not sufficient for visual awareness.
2001, 11:966-974. Neuropsychologia 2008, 46:864-869.
11. Nakamura K, Kuo WJ, Pegado F, Cohen L, Tzeng OJ, Dehaene S: 26. Kentridge RW, Heywood CA, Weiskrantz L: Spatial attention
Universal brain systems for recognizing word shapes and speeds discrimination without awareness in blindsight.
handwriting gestures during reading. Proc Natl Acad Sci U S A Neuropsychologia 2004, 42:831-835.
2012, 109:20762-20767.
27. Naccache L, Blandin E, Dehaene S: Unconscious masked
12. Qiao E, Vinckier F, Szwed M, Naccache L, Valabregue R, priming depends on temporal attention. Psychol Sci 2002,
 Dehaene S, Cohen L: Unconsciously deciphering handwriting: 13:416-424.
subliminal invariance for handwritten words in the visual word
form area. Neuroimage 2010, 49:1786-1799. 28. Wyart V, Tallon-Baudry C: Neural dissociation between
This study uses subliminal priming and functional MRI to prove that even visual awareness and spatial attention. J Neurosci 2008,
hard-to-read handwritten words cause repetition enhancement in the 28:2667-2679.
visual word form area, suggesting that invisible visual stimuli are pro-
cessed non-consciously up to a high level of perceptual invariance. 29. Wyart V, Dehaene S, Tallon-Baudry C: Early dissociation
between neural signatures of endogenous spatial attention
13. Van Opstal F, de Lange FP, Dehaene S: Rapid parallel semantic and perceptual awareness during visual masking. Front Hum
processing of numbers without awareness. Cognition 2011, Neurosci 2012, 16:1-14.
120(1):136-147.
30. Watanabe M, Cheng K, Murayama Y, Ueno K, Asamizuya T,
14. Luck SJ, Vogel EK, Shapiro KL: Word meanings can be Tanaka K, Logothetis N: Attention but not awareness modulates
accessed but not reported during the attentional blink. Nature the BOLD signal in the human V1 during binocular
1996, 383:616-618. suppression. Science 2011, 334:829-831.
15. Devlin JT, Jamison HL, Matthews PM, Gonnerman LM: 31. Rahnev D, Maniscalco B, Graves T, Huang E, de Lange FP, Lau H:
Morphology and the internal structure of words. Proc Natl Acad  Attention induces conservative subjective biases in visual
Sci U S A 2004, 101:14984-14988. perception. Nat Neurosci 2011, 14:1513-1515.
When attended and unattended visual stimuli were matched for objective
16. Gaillard R, Del Cul A, Naccache L, Vinckier F, Cohen L, Dehaene S: performance, the unattended stimuli were paradoxically rated as
 Nonconscious semantic processing of emotional words being more visible than the attended ones on a subjective visibility task.
modulates conscious access. Proc Natl Acad Sci U S A 2006, This study therefore provides one of the clearest piece of evidence for a
103:7524-7529. dissociation between attention and conscious perception.
Using subliminal words, this study shows that, during masking, the
threshold for conscious perception is lowered firstly, for masked words 32. Sergent C, Wyart V, Babo-Rebelo M, Cohen L, Naccache L,
that convey an emotional meaning (e.g. rape) compared to words that  Tallon-Baudry C: Cueing attention after the stimulus is gone
convey a neutral meaning (e.g. cape); and secondly, for masked words can retrospectively trigger conscious perception. Curr Biol
that have been previously seen, compared to words that have never been 2013, 23:150-155.
seen before. Thus, subliminal processing is modulated by both emotion This study shows that a visual stimulus presented at threshold shows
and prior knowledge. enhanced detection when it is ‘post-cued’ by a subsequent attentional

www.sciencedirect.com Current Opinion in Neurobiology 2014, 25:76–84

82 Theoretical and computational neuroscience

cue presented several hundreds of milliseconds after the initial stimulus. 44. Dupoux E, de Gardelle V, Kouider S: Subliminal speech
The authors conclude that ‘this retroperception effect demonstrates that perception and auditory streaming. Cognition 2008,
postcued attention can retrospectively trigger the conscious perception 109:267-273.
of a stimulus that would otherwise have escaped consciousness’.
45. Lau H, Rosenthal D: Empirical support for higher-order theories
33. van Gaal S, Scholte HS, Lamme VA, Fahrenfort JJ, of conscious awareness. Trends Cogn Sci 2011, 15:365-373.
Ridderinkhof KR: Pre-SMA graymatter density predicts
individual differences in action selection in the face of 46. Charles L, Van Opstal F, Marti S, Dehaene S: Distinct brain
conscious and unconscious response conflict. J Cogn  mechanisms for conscious versus subliminal error detection.
Neurosci 2010, 23:382-390. Neuroimage 2013, 73:80-94.
This study indicates a dissociation between two metacognitive mechan-
34. van Gaal S, Ridderinkhof KR, Scholte HS, Lamme VA: isms of error detection. When performing a forced-choice task with
Unconscious activation of the prefrontal no-go network. masked stimuli, subjects performed above chance level in guessing
J Neurosci 2010, 30:4143-4150. whether they made an error or not, even when the stimulus was sub-
jectively invisible. Contrasting with this graded non-conscious metacog-
35. van Gaal S, Lamme VA, Fahrenfort JJ, Ridderinkhof KR: nitive performance, however, the error-related negative behaved in an all-
 Dissociable brain mechanisms underlying the conscious and or-none manner and was only present on visible trials. The authors
unconscious control of behavior. J Cogn Neurosci 2010, conclude that metacognition can partially operate non-consciously,
23:91-105. but that consciousness is requiredin order to achievea near-certainty
In this experiment, metacontrast masking was used to present subliminal of having made an error.
stop signals. These non-conscious signals slowed down the participants’
responses and strongly modulated early electrophysiological markers of 47. Kanai R, Walsh V, Tseng CH: Subjective discriminability of
frontal inhibitory control (N2). However, the subsequent P3 event was invisibility: a framework for distinguishing perceptual and
much larger on conscious than on non-conscious trials. Together with attentional failures of awareness. Conscious Cogn 2010,
Refs. [34,36], this research suggests that part, but not all of the prefrontal 19(4):1045-1057.
mechanisms of inhibitory control can be activated non-consciously.
48. Gaillard R, Dehaene S, Adam C, Clemenceau S, Hasboun D,
36. van Gaal S, Ridderinkhof KR, Fahrenfort JJ, Scholte HS,  Baulac M, Cohen L, Naccache L: Converging intracranial
Lamme VA: Frontal cortex mediates unconsciously triggered markers of conscious access. PLoS Biol 2009, 7:e61.
inhibitory control. J Neurosci 2008, 28:8053-8062. Using intracranial recordings in 10 epilepsy patients during a masking
paradigm, the authors report that subliminal stimuli evoked easily detect-
37. Nieuwenhuis S, Ridderinkhof KR, Blom J, Band GP, Kok A: Error- able early event-related potentials and gamma-band activity that propa-
 related brain potentials are differentially related to awareness gate in a feed-forward manner with the first few hundred milliseconds
of response errors: evidence from an antisaccade task. after stimulus onset. Only conscious stimuli, however, evoked late and
Psychophysiology 2001, 38:752-760. long-lasting potentials over frontal electrodes as well as back into the
This classic study used an occulomotor anti-saccade task to show that visual cortex. Conscious perception was also associated with strong and
the error-related negativity (ERN), a known brain marker of error detec- late increases in gamma-band power, long-distance beta synchrony and
tion, remained present on trials in which subjects failed to consciously bidirectional Grainger causality. All of these markers may index the late
detect their eye-movement errors. Only the later Pe component varied ignition of a global neuronal workspace associated with conscious
with subjective error awareness. Thus, performance monitoring can processing.
occur in the absence of conscious perception.
49. Fisch L, Privman E, Ramot M, Harel M, Nir Y, Kipervasser S,
38. Logan GD, Crump MJ: Cognitive illusions of authorship reveal  Andelman F, Neufeld MY, Kramer U, Fried I et al.: Neural
 hierarchical error detection in skilled typists. Science 2010, ‘‘ignition’’: enhanced activation linked to perceptual
330:683-686. awareness in human ventral stream visual cortex. Neuron
While skilled typists were asked to type a target word, a computer 2009, 64:562-574.
program surreptitiously inserted errors or, conversely, corrected the Using electro-corticograms (ECOG) in 11 epilepsy patients, the authors
errors made by the subjects. Subjects often did not detect these manip- identified several occipito-temporal sites where gamma-band power was
ulations and claimed responsibility for the words as they appeared on selectively evoked by one category of visual stimuli (e.g. pictures of
screen. However, their typing rates revealed post-error slowing after faces). When the stimuli were masked near the visibility threshold, only
genuine errors, even when they were computer-corrected and therefore the pictures reported as visible led to a sharp and long-lasting ‘ignition’, a
were not detected consciously, but not after pseudo-errors inserted by strong increase in gamma-band power. Invisible stimuli only led to a
the program. These findings support a hierarchical model of error detec- modest and short-lived gamma burst (see Figure 6c). Thus, gamma-band
tion, in which action monitoring may partially unfold non-consciously. ignition, particularly after 200–300 ms, is a candidate marker of conscious
perception.
39. Lau HC, Passingham RE: Unconscious activation of the
 cognitive control system in the human prefrontal cortex. J 50. Aru J, Axmacher N, Do Lam AT, Fell J, Elger CE, Singer W,
Neurosci 2007, 27:5805-5811. Melloni L: Local category-specific gamma band responses in
The authors used subliminal priming to evaluate if task switching could be the visual cortex do not reflect conscious perception. J
partially launched without awareness. When presented with subliminal Neurosci 2012, 32:14909-14914.
cues coding for a switch to the alternative task, subjects were less
accurate in performing the current task, and activity in regions associated 51. van Gaal S, Lamme VA: Unconscious high-level information
with the current task decreased while activity in regions associated with processing: implication for neurobiological theories of
the alternative task increased. This study therefore suggests that sub- consciousness. Neuroscientist 2012, 18:287-301.
liminal cues can reach higher-order stages of cognitive control.
52. Lamme VA: Towards a true neural stance on consciousness.
40. Reuss H, Kiesel A, Kunde W, Hommel B: Unconscious activation Trends Cogn Sci 2006, 10:494-501.
of task sets. Conscious Cogn 2011, 20:556-567.
53. Tononi G: An information integration theory of consciousness.
41. Soto D, Mantyla T, Silvanto J: Working memory without BMC Neurosci 2004, 5:42.
 consciousness. Curr Biol 2011, 21:R912-R913.
This provocative study shows that an undetected masked grating, rated 54. Tononi G: Consciousness as integrated information: a
as subjectively invisible, may nevertheless lead to better-than-chance provisional manifesto. Biol Bull 2008, 215:216-242.
performance in a delayed match-to-sample task. Working memory per-
formance remained above even when the sample appeared after a delay 55. Cohen MX, van Gaal S, Ridderinkhof KR, Lamme VA:
of 5 s, and even when the delay period was briefly interrupted by an Unconscious errors enhance prefrontal-occipital oscillatory
irrelevant distractor. The authors conclude that working memory can synchrony. Front Hum Neurosci 2009, 3:54.
partially operate without conscious perception.
56. Pitts MA, Martinez A, Hillyard SA: Visual processing of contour
42. Mattler U: Inhibition and decay of motor and nonmotor priming.  patterns under conditions of inattentional blindness. J Cogn
Percept Psychophys 2005, 67:285-300. Neurosci 2012, 24:287-303.
This careful event-related potential study uses inattentional blindness to
43. Greenwald AG, Draine SC, Abrams RL: Three cognitive markers present above-threshold stimuli with or without conscious perception.
of unconscious semantic activation. Science 1996, The authors also attempt to separate conscious perception from further
273:1699-1702. stimulus processing by making the perceived stimulus task-relevant or

Current Opinion in Neurobiology 2014, 25:76–84 www.sciencedirect.com

 Theorizing conscious processing Dehaene et al. 83

task-irrelevant on distinct blocks. The findings suggest that conscious task with masked stimuli. Subjects rated the subjective visibility of each
perception per secorrelates with a late posterior negativity. A P300 target (‘clear’ or ‘unclear’). TMS reduced subjective visibility and second-
component only appears when the visual stimulus is both seen and order metacognitive d-prime, while first-order performance remained unaf-
task-relevant. fected. Thus, prefrontal cortex seems to play a prominent role in second-
order subjective visibility judgments, even for simple visual stimuli.
57. Mudrik L, Breska A, Lamy D, Deouell LY: Integration without
awareness. Psychol Sci 2011, 22:764-770. 74. Quiroga RQ, Mukamel R, Isham EA, Malach R, Fried I: Human
single-neuron responses at the threshold of conscious
58. de Lange FP, van Gaal S, Lamme VA, Dehaene S: How awareness recognition. Proc Natl Acad Sci U S A 2008, 105:
changes the relative weights of evidence during human 3599-3604.
decision-making. PLoS Biol 2011, 9:e203-e1001.
75. de Gardelle V, Charles L, Kouider S: Perceptual awareness and
59. Vorberg D, Mattler U, Heinecke A, Schmidt T, Schwarzbach J: categorical representation of faces: evidence from masked
Different time courses for visual perception and action priming. Conscious Cogn 2011, 20:1272-1281.
priming. Proc Natl Acad Sci U S A 2003, 100:6275-6280.
76. de Gardelle V, Kouider S, Sackur J: An oblique illusion
60. Del Cul A, Baillet S, Dehaene S: Brain dynamics underlying the modulated by visibility: non-monotonic sensory integration in
nonlinear threshold for access to consciousness. PLoS Biol orientation processing. J Vis 2010, 10:6.
2007, 5:e260.
77. Melloni L, Molina C, Pena M, Torres D, Singer W, Rodriguez E:
61. Melloni L, Schwiedrzik CM, Muller N, Rodriguez E, Singer W: Synchronization of neural activity across cortical areas
 Expectations change the signatures, timing of correlates with conscious perception. J Neurosci 2007,
electrophysiological correlates of perceptual awareness. J 27:2858-2865.
Neurosci 2011, 31:1386-1396.
This important event-related potential study extends an earlier study by 78. Woodman GF: Masked targets trigger event-related potentials
Del Cul et al. (2007, Ref. [60]). It replicates the finding that conscious visual indexing shifts of attention but not error detection.
perception of brief unexpected stimuli is associated primarily with a late Psychophysiology 2010, 47(3):410-414.
P300 potential, but also shows that the neural correlates of conscious
perception shift to an earlier time period (the P2) when the identity of the 79. Wyart V, Tallon-Baudry C: How ongoing fluctuations in
stimulus is fully expected. human visual cortex predict perceptual awareness:
baseline shift versus decision bias. J Neurosci 2009,
62. Dehaene S: Conscious and nonconscious processes: distinct 29:8715-8725.
forms of evidence accumulation? In Better Than Conscious?
Decision Making, the Human Mind, and Implications For 80. Summerfield C, Egner T, Greene M, Koechlin E, Mangels J,
Institutions. Strüngmann Forum Report. Edited by Engel C, Singer Hirsch J: Predictive codes for forthcoming perception in the
W. MIT Press; 2008. frontal cortex. Science 2006, 314:1311-1314.
63. Shadlen MN: Consciousness as a decision to engage. In 81. Kouider S, Stahlhut C, Gelskov SV, Barbosa L, Dutat M, de
Characterizing Consciousness: From Cognition to the Clinic?. Gardelle V, Christophe A, Dehaene S, Dehaene-Lambertz G: A
Edited by Dehaene S, Christen Y.. Research and Perspectives in neural marker of perceptual consciousness in infants. Science
NeurosciencesSpringer Verlag; 2011:27-46. 2013, 340(6130):376-380.
82. Aru J, Bachmann T, Singer W, Melloni L: Distilling the neural
64. Lau HC: A higher order Bayesian decision theory of  correlates of consciousness. Neurosci Biobehav Rev 2012,
consciousness. Prog Brain Res 2007, 168:35-48. 36:737-746.
65. Ko Y, Lau H: A detection theoretic explanation of blindsight This position paper contains an important critique of the contrastive
suggests a link between conscious perception and method, used in most empirical studies of conscious processing, and
metacognition. Philos Trans R Soc Lond B: Biol Sci 2012, which consists in contrasting behavior and brain activity evoked by
367:1401-1411. conscious versus non-conscious stimuli. The authors argue that the
differences between conscious and non-conscious conditions may not
66. Pasquali A, Timmermans B, Cleeremans A: Know thyself: solely reflect the genuine neural correlates of conscious processing, but
metacognitive networks and measures of consciousness. also include the ‘prerequisites for and consequences of conscious
Cognition 2010, 117:182-190. processing of the particular content’.

67. Wong KF, Wang XJ: A recurrent network mechanism of time 83. Railo H, Koivisto M: The electrophysiological correlates of
integration in perceptual decisions. J Neurosci 2006, stimulus visibility and metacontrast masking. Conscious Cogn
26:1314-1328. 2009, 18(3):794-803.
68. Dehaene S, Sergent C, Changeux JP: A neuronal network model 84. Wacongne C, Labyt E, van Wassenhove V, Bekinschtein T,
linking subjective reports and objective physiological data Naccache L, Dehaene S: Evidence for a hierarchy of predictions
during conscious perception. Proc Natl Acad Sci U S A 2003, and prediction errors in human cortex. Proc Natl Acad Sci U S A
100:8520-8525. 2011, 108:20754-20759.
69. Dehaene S, Changeux JP: Ongoing spontaneous activity 85. King JR, Faugeras F, Gramfort A, Schurger A, El Karoui I, Sitt JD,
controls access to consciousness: a neuronal model for Wacongne C, Labyt E, Bekinschtein P, Naccache L et al.: Single-
inattentional blindness. PLoS Biol 2005, 3:e141. trial decoding of auditory novelty responses facilitates the
detection of residual consciousness. Neuroimage 2013,
70. Dehaene S, Changeux JP: Experimental and theoretical 83:726-738.
approaches to conscious processing. Neuron 2011,
70:200-227. 86. Shanahan M, Baars B: Applying global workspace theory to the
frame problem. Cognition 2005, 98:157-176.
71. Sergent C, Baillet S, Dehaene S: Timing of the brain events
underlying access to consciousness during the attentional 87. Dehaene S, Kerszberg M, Changeux JP: A neuronal model of a
blink. Nat Neurosci 2005, 8:1391-1400. global workspace in effortful cognitive tasks. Proc Natl Acad
Sci U S A 1998, 95:14529-14534.
72. Schwiedrzik CM, Ruff CC, Lazar A, Leitner FC, Singer W, Melloni L:
Untangling perceptual memory: hysteresis and adaptation 88. van den Heuvel MP, Kahn RS, Goni J, Sporns O: High-cost, high-
map into separate cortical networks. Cereb Cortex 2012. [Epub capacity backbone for global brain communication. Proc Natl
ahead of print]. Acad Sci U S A 2012, 109(28):11372-11377.
73. Rounis E, Maniscalco B, Rothwell JC, Passingham R, Lau H: 89. van den Heuvel MP, Sporns O: Rich-club organization
 Theta-burst transcranial magnetic stimulation to the of the human connectome. J Neurosci 2011,
prefrontal cortex impairs metacognitive visual awareness. 31:15775-15786.
Cogn Neurosci 2010.
Repetitive transcranial magnetic stimulation was applied to bilateral 90. Sackur J, Dehaene S: The cognitive architecture for chaining of
dorsolateral prefrontal cortex while subjects performed a discrimination two mental operations. Cognition 2009, 111:187-211.

www.sciencedirect.com Current Opinion in Neurobiology 2014, 25:76–84

84 Theoretical and computational neuroscience

91. Fan Z, Singh K, Muthukumaraswamy S, Sigman M, Dehaene S, 105. Seth AK, Izhikevich E, Reeke GN, Edelman GM: Theories and
Shapiro K: The cost of serially chaining two cognitive measures of consciousness: an extended framework. Proc
operations. Psychol Res 2012, 76:566-578. Natl Acad Sci U S A 2006, 103:10799-10804.
92. de Lange FP, Jensen O, Dehaene S: Accumulation of evidence 106 Lewis LD, Weiner VS, Mukamel EA, Donoghue JA, Eskandar EN,
during sequential decision making: the importance of top-  Madsen JR, Anderson WS, Hochberg LR, Cash SS, Brown EN
down factors. J Neurosci 2010, 30:731-738. et al.: Rapid fragmentation of neuronal networks at the onset
of propofol-induced unconsciousness. Proc Natl Acad Sci U S
93. Schwiedrzik CM, Singer W, Melloni L: Subjective and objective A 2012, 109:E3377-E3386.
learning effects dissociate in space and in time. Proc Natl Acad In this rare study, single-neuron recordings were obtained while three
Sci U S A 2011, 108:4506-4511. epilepsy patients underwent propofol anesthesia. The loss of conscious-
94. Marti S, Sigman M, Dehaene S: A shared cortical bottleneck ness was associated with a sudden onset of slow oscillations (<1 Hz) in
underlying Attentional Blink and Psychological Refractory the local field potential. Local neuronal networks continued to exhibit
Period. Neuroimage 2012, 59:2883-2898. spiking activity, but this activity became fragmented in time by the slow
oscillations, and decorrelated in space as the oscillations were asyn-
95. Marti S, Sackur J, Sigman M, Dehaene S: Mapping chronous across distant sites. The suggestion is that temporal and spatial
 introspection’s blind spot: reconstruction of dual-task disintegration of cortical activity may cause a loss of consciousness.
phenomenology using quantified introspection. Cognition
2010, 115:303-313. 107. Schroter MS, Spoormaker VI, Schorer A, Wohlschlager A,
Although introspection is usually considered with much suspicion in Czisch M, Kochs EF, Zimmer C, Hemmer B, Schneider G, Jordan D
cognitive psychology, this study shows that human subjects can accu- et al.: Spatiotemporal reconfiguration of large-scale brain
rately report on the trial-by-trial variations in their own processing time. functional networks during propofol-induced loss of
They only suffer from a specific illusion: in a dual-task setting, when they consciousness. J Neurosci 2012, 32:12832-12840.
are distracted by a primary task, their introspection of the onset of a
second stimulus is delayed and diffused — a phenomenon analogous to 108. Supp GG, Siegel M, Hipp JF, Engel AK: Cortical hypersynchrony
the attentional blink. predicts breakdown of sensory processing during loss of
consciousness. Curr Biol 2011, 21:1988-1993.
96. Asplund CL, Todd JJ, Snyder AP, Gilbert CM, Marois R: Surprise-
induced blindness: a stimulus-driven attentional limit to 109. Vijayan S, Ching S, Purdon PL, Brown EN, Kopell NJ:
conscious perception. J Exp Psychol Hum Percept Perform Thalamocortical mechanisms for the anteriorization of alpha
2011, 36:1372-1381. rhythms during propofol-induced unconsciousness. J
Neurosci 2013, 33:11070-11075.
97. Nieuwenstein MR, Potter MC, Theeuwes J: Unmasking the
attentional blink. J Exp Psychol Hum Percept Perform 2009, 110. Ching S, Cimenser A, Purdon PL, Brown EN, Kopell NJ:
35:159-169. Thalamocortical model for a propofol-induced alpha-rhythm
associated with loss of consciousness. Proc Natl Acad Sci U S
98. Tombu MN, Asplund CL, Dux PE, Godwin D, Martin JW, Marois R: A 2010, 107:22665-22670.
A unified attentional bottleneck in the human brain. Proc Natl
Acad Sci U S A 2011, 108:13426-13431. 111. Owen AM, Coleman MR, Boly M, Davis MH, Laureys S, Pickard JD:
Detecting awareness in the vegetative state. Science 2006,
99. Zylberberg A, Fernandez Slezak D, Roelfsema PR, Dehaene S, 313:1402.
Sigman M: The brain’s router: a cortical network model of
serial processing in the primate brain. PLoS Comput Biol 2010, 112 Casali A, Gosseries O, Rosanova M, Boly M, Sarasso S,
6:e765.  Casali KR, Casarotto S, Bruno MA, Laureys S, Tononi G et al.: A
theoretically based index of consciousness independent of
100. Schurger A, Pereira F, Treisman A, Cohen JD: Reproducibility sensory processing and behavior. Sci Transl Med 2013,
distinguishes conscious from nonconscious neural 5:198ra105.
representations. Science 2009, 327:97-99. This study proposes a quantitative assesmentof the state of conscious-
101. Chennu S, Noreika V, Gueorguiev D, Blenkmann A, Kochen S, ness. By quantifying the complexity of the event-related response to a
Ibanez A, Owen AM, Bekinschtein TA: Expectation and attention TMS pulse, the authors obtain a numerical index that shows a complete
in hierarchical auditory prediction. J Neurosci 2013, 33: bimodal separation of conscious and non-conscious normal subjects
11194-11205. (during sleep or anesthesia). When applied to a small group of patients
with disorders of consciousness, this index also separates vegetative and
102. Bekinschtein TA, Dehaene S, Rohaut B, Tadel F, Cohen L, minimally conscious patients.
Naccache L: Neural signature of the conscious processing of
auditory regularities. Proc Natl Acad Sci U S A 2009, 113 King JR, Sitt JD, Faugeras F, Rohaut B, El Karoui I, Cohen L,
106:1672-1677.  Naccache L, Dehaene S: Information sharing in the brain
indexes consciousness in non-communicative patients. Curr
103. Fahrenfort JJ, Snijders TM, Heinen K, van Gaal S, Scholte HS, Biol 2013, 23:1914-1919.
Lamme VA: Neuronal integration in visual cortex elevates face This study introduces a novel mathematical measure, weighted symbolic
category tuning to conscious face perception. Proc Natl Acad mutual information, which can accurately quantify the amount of informa-
Sci U S A 2012, 109:21504-21509. tion exchanged between two recording sites. Applied to a large set of
high-resolution EEG recordings from patients with disorders of con-
104. Barrett AB, Seth AK: Practical measures of integrated sciousness, this measure separates vegetative and minimally conscious
information for time-series data. PLoS Comput Biol 2011, patients, regardless of etiology and delay since insult.
7:e0521001.

Current Opinion in Neurobiology 2014, 25:76–84 www.sciencedirect.com