Decoding the Neural Signature of Consciousness

Consciousness has kept philosophers and scientists occupied for centuries. Lofty ideas related to humanity, agency, and responsibility all relate to the thing we call “consciousness”; and yet, we still don’t understand how this elevated concept plays out on a mechanistic level within individual people. Is it possible to pinpoint when and how conscious awareness occurs?

According to psychological scientist Stanislas Dehaene (Collège de France and INSERM-CEA Cognitive Neuroimaging Unit, France), the answer may be yes. In his  keynote address at the International Convention of Psychological Science in Amsterdam, the Netherlands, Dehaene showed how the advent of new tools is bringing us closer than ever to cracking the neural code of consciousness.

An essential component of cracking the code is being able to identify and follow stages of processing in the brain. In the 1800s, the Dutch scientist Franciscus Donders showed that our responses to environmental stimuli aren’t instantaneous — it takes time for us to sense a stimulus, become aware of it, and then formulate and execute a response. While this process typically happens within the span of half a second or less, such quick reactions belie significant mental processing.

Ever since the days of Donders, psychological scientists have been working toward unpacking mental processing. Being able to decompose mental operations into their various stages helps us understand not only how we process things but also why a particular sequence of processing leads, or doesn’t lead, to conscious awareness.

Unfortunately, our standard methods of investigation don’t provide much help with this unpacking. Measuring reaction time may provide a basic tool for gauging the effort involved in mental processing, and brain imaging tools can tell us the areas of the brain involved in processing a given stimulus — but neither approach provides enough information to illuminate exactly how and why we respond the way we do.

But if we combine brain imaging tools that have high temporal resolution — such as magnetoencephalography (MEG) or electroencephalography (EEG) — with machine learning programs, Dehaene noted, we can finally break information processing down and watch it unfold in real time.

Using a mathematical technique called “multivariate decoding,” machine learning programs can identify patterns of brain activity and learn to associate those patterns in time with the presentation of a stimulus. The program can then identify which MEG or EEG sensors are able to “recover” the information the participant is processing at a given time.

With such highly sensitive tools, we are not limited to asking where activation is occurring; we can obtain a much more detailed picture of how individual responding relates, in psychological terms, to the stimulus.

In other words, said Dehaene, we can “return to the stimulus space and ask, What was the identity of the stimulus? Where was it on screen? Did the subject register that it was a digit or not? And so on.”

Decoding can tell us, for example, whether processing is fleeting or sustained, whether it happens quickly or slowly, and even whether information is reactivated later.

And the technique conveys another advantage in that it can be applied to MEG or EEG data from individual participants, thereby avoiding the smoothing out of differences in individual responses that occurs when data are averaged prior to analysis.

With this decoding method, Dehaene and colleagues have been able to demonstrate a clear dissociation between two similar neurological responses that are associated with novelty — mismatch negativity and the P300 brain wave. The dissociation is evident even in the absence of an overt behavioral response, and work in collaboration with Lionel Naccache of the ICM Brain and Spine Institute in Paris has shown that the presence of P300 serves as a marker of conscious processing, while mismatch negativity does not.

These findings suggest that the ability to decode P300 responses could provide a useful clinical tool to help distinguish, for example, whether a patient is minimally conscious or in a persistent vegetative state. And although the P300 is just one potential marker of consciousness, work in this area has helped to delineate the aspects of processing that are unconscious and those that are conscious, informing an overarching theory of consciousness.

According to Dehaene’s “global neuronal workspace” theory, consciousness corresponds to a system that breaks the modularity of local circuits, enabling the circuits to talk to each other and share information. When there is enough information to surpass a certain threshold, the information reverberates through the highly interconnected system, gaining access to the so-called global workspace — a process termed “ignition.”

Experiments using a masking paradigm with varying lengths of delay show this ignition phenomenon clearly, even in infants: Decoding early processing produces linear patterns of activation, but at a certain point the decoder suddenly produces a nonlinear pattern, indicating ignition.

Thus, what we experience as consciousness, Dehaene explained, “is actually a global availability of information.”

“This is perhaps the best picture we have of conscious processing so far,” Dehaene said, but it still leaves many open questions. For example, although these methods allow us to see processing of information as it unfolds, isolating the exact moment when information transitions from unconsciousness to consciousness is a much more difficult task.

Many labs are now using multivariate decoding to tackle these open questions, and Dehaene encouraged the members of the audience to join in on the project, noting that free and open-source code for processing MEG and EEG data is available.

Dehaene’s most recent book, Consciousness and the Brain: Deciphering How the Brain Codes Our Thoughts, provides further insight into his work on consciousness.