I don't usually do hot off the press reports but this one demanded my attention. Consciousness is one of the greatest unsolved mysteries of modern science.
I report today (2009) on the paper Converging Intracranial Markers of Conscious Access by Gaillard, Dehaene, Adam, Clemenceau, Hasboun, Baulac, Cohen, and Naccache that appeared in the March 2009 issue of PLoS Biology, vol 7, issue 3. (By the way I commend the PLoS series of publications to you. This is the Public Library of Science, which publishes its articles and pdf's on-line immediately for free. The era of keeping science locked behind closed doors is almost over. Taxpayers: you should demand it!)
This paper immediately struck me as important, because 1) it reports key, possibly defining neural correlates of consciousness (NCC’s); 2) the researchers are a respected European group with long experience in this field; 3) the work was done in ten human beings who had craniotomies and had dense intracranial electrode arrays placed on the surface of their brains .; 4) this is a sophisticated comparison of the anatomy and physiology of consciousness in real-time as the processes unfold. 5) Most important - these findings appear to confirm the leading hypothesis in the field of neural consciousness research: global workspace theory.
The research set up is succinctly described in this editorial from PLoS by Richard Robinson . The goal of the study procedure was to compare a conscious experience to a similar nonconscious (non?)experience.
So - ten human subjects (post craniotomy) with implanted electrode arrays ... Each subject was shown a printed word on a computer screen that was either UNMASKED and hence VISIBLE or MASKED and hence INVISIBLE.
Here's the set-up.
First (in every presentation, both masked and unmasked) a row of hatch marks (#########) was shown on the screen as a meaningless mask.
Next, a printed word like cousin was flashed on the screen very briefly (for only 29 milliseconds (msec)) – a thirtieth of a second.
Next, to make the word invisible during the MASKED presentation, a row of ampersands would be flashed for 400 msec .
The effect of this MASK was to retroactively interfere with the neural processing of the word. With masking, the subjects could not guess the words or even the gist of the words (whether they were threatening words like kill or danger or non-threatening words like cousin or see.) (However, note, that the printed word would still have appeared on the subject's retinas and caused something to happen in their visual systems, even though they could report nothing.)
The other half of the time, rather than being masked by the , the printed word would again be presented for 29 msec, but then would simply be followed by a blank screen (no mask). Since there was nothing to interfere or mask the visual experience of the printed word, the subjects were able to say the printed word that they saw - they had actually seen it.
So, the million dollar question was this... what was going on in the brains of the patients when they actually SAW the unmasked or visible printed word (in comparison to the same word when it was masked).
The above figure compares unconscious responses on the left to conscious responses on the right. In the conscious column (right-hand) note the intense, high-frequency responses that begin around 200 msec and last for another 400 msec or so. That is the signature of global resonance.
Here's the bottom line, and I've simplified it for this summary.
Both the masked words and the unmasked words excited the early stages of visual processing in the occipital lobe (back of the brain). After all, even the masked words were detected, ie, processed by the retina. However, just that early, brief processing in the occipital lobe was not sufficient to give rise to the subjective experience of seeing and identifying the word.
For consciousness to occur, what had to happen was this. The initial wave of visual processing of the printed word in occipital (and then temporal lobe) had to be sufficiently long and intense to excite neurons in the prefrontal cortex (front of the head).
Once prefrontal cortex was excited, it would then cause a wave of feedback to the occipital and temporal lobes that would amplify their excitation. This would result in a long-lasting reverberation between the the front of the brain and the back of the brain . This, then, appears to be the defining characteristic of consciousness: widespread synchronous activation of many brain regions particularly the visual cortex (for vision) plus the prefrontal cortex (the I part of I see the word or the world.)
This widespread activation of many brain regions is the defining characteristic of the GLOBAL WORKSPACE THEORY OF CONSCIOUSNESS .
I first became acquainted with the precursors of this theory in the 1970's when it was known as the BLACKBOARD system or blackboard architecture. It originated in the work of cognitive and AI scientists Herb Simon (Nobel laureate) and Allen Newell at Carnegie Mellon and was also used in AI systems for speech recognition.
The theory was picked up by cognitive scientist Bernard Baars and applied to consciousness where it became known as global workspace theory.
The essence of the theory is this. Consciousness is a diffuse, global process that broadcasts a model of the subject's world to all four lobes of the cortex. In particular, in addition to whatever primary sensory cortex is involved (vision - occipital; hearing - temporal; touch - parietal), diffuse temporoparietal association areas are excited as well as (and especially) prefrontal cortex. This widespread activation of neural networks is the global workspace or blackboard of the mind.
Here, in the words of the authors is the essence of the global workspace model:
“many modular cerebral networks are active in parallel and process information in an unconscious manner. However incoming visual information only becomes conscious if and only if the three following conditions are met :
Condition 1: information must be explicitly represented by the neuronal ﬁring of perceptual networks located in visual cortical areas coding for the speciﬁc features of the conscious percept.
Condition 2: this neuronal representation must reach a minimal threshold of duration and intensity necessary for access to a second stage of processing, associated with a distributed cortical network involved in particular parietal and prefrontal cortices.
Condition 3: through joint bottom-up propagation and top-down attentional ampliﬁcation, the ensuing brain-scale neural assembly must ‘‘ignite’’ into a self-sustained reverberant state of coherent activity that involves many neurons distributed throughout the brain .
The data that they uncovered precisely supports this theory. The key difference that they showed between the neural response to the masked words and the unmasked or visible (consciously seen) words was this:
Nonconscious processing of words elicited short-lasting activity across multiple cortical areas, including parietal and visual areas. In sharp contrast, only consciously perceived words were accompanied by long-lasting effects (200 ms) across a great variety of cortical sites, with a special involvement of the prefrontal lobes . This sustained pattern of neural activity was characterized by a specific increase of coherence between distant areas, suggesting conscious perception is broadcasted widely across the cortex. In particular, when their human subjects actually saw the words (unmasked) their neural responses showed sustained voltage changes especially in the prefrontal cortex (the executive area of the brain) and large increases in gamma power. (Gamma is the highest frequency band on the eeg (above 30 hz ) and is a well-known hallmark of conscious awareness.) Their last two findings - long-distance phase synchrony and long-distance Granger causality - are also important, since they show that all the areas connected are actually communicating information. They are not only in-phase but causally connected. Here are some relevant quotes.
These observations ﬁt with the global workspace model, which postulates that once a representation is consciously accessed, a broad distributed network, involving in particular prefrontal cortex, ignites and broadcasts its content. Computer simulations of this ignition process within thalamo-cortical networks have shown strong recurrent interactions occuring exclusively during conscious access, implying that the relevant neurons are suddenly coactivated in a cooperative manner and emit increased and synchronized high-frequency oscillations in the gamma range.
These results support our prediction that conscious access involves an improved long-distance exchange of information across a very broad cortical network. Long-distance coherence between distant regions appears to co-occur with the active maintenance of locally coded representations. ...
These observations are consistent with global workspace theory’s prediction that conscious access occurs during a late time period and is dominated by an in-ﬂow of perceptual information from posterior visual areas into higher frontal and parietal cortices.
A simplified explanation is that the areas where the detailed encoding of visual information occurs in the back of the brain are excited in a sustained manner by the areas which they excite in the front of the brain (prefrontal cortex). The widespread reverberation that ensues is an essential neural correlate of consciousness (NCC).
These findings are moving us toward a solution to the “easy problem” of consciousness – what are its neural correlates. The hard problem remains – how does a diffuse, oscillating network of neurons cause subjective experience. (See the Mystery of Consciousness .)