Revelations from the Unconscious

GROWTH DESPITE OBSTACLES

Only a handful of labs worldwide use brain imaging to explore consciousness disorders. The goal is to determine the activity patterns missing in PVS or MCS that, when present, enable someone to be conscious. Progress has been slowed by several factors: inadequate funding; practical problems such as unconscious patients' inability to stay immobile during brain scans; and ethical objections to experiments on these nonconsenting subjects. But investigators also point to signs of growth. "Five years ago, it seemed as though there was almost nobody working in this area," recalls Owen. Now, he says he sees Laureys two or three times a year at meetings where consciousness is discussed.

Nevertheless, doubts remain about imaging's potential to elucidate consciousness. Christof Koch, a professor of cognitive and behavioral biology at the California Institute of Technology, cautions that brain scans detect hemodynamic, not neuronal, activity (though a new paper correlates fMRI with neuronal firing²). Imaging, moreover, does not show what people perceive, notes Jordan Grafman, chief of the cognitive neuroscience section at the National Institute of Neurological Disorders and Stroke. "In the end, that's consciousness," Grafman states. "That's what you want to know."

Stanislas Dehaene, director of the cognitive neuroimaging unit of the Service Hospitalier Frédéric Joliot in Orsay, France, has a different perspective. He studies how conscious subjects subliminally process camouflaged, or "masked," stimuli. Asserting that an "enormous amount" can be learned from patients with consciousness disorders, he explains: "The crossing of the threshold of consciousness in a masking situation... is a large event. So is the recovery of consciousness in vegetative-state patients." These different paradigms, he continues, "converge to the same style of brain activation," involving the cortex's parietal, prefrontal, and cingulate network.

POIGNANT E-MAILS

Coma, PVS, and MCS are clinical diagnoses that are each compatible with a variety of brain damage. After anoxia or traumatic brain injury, patients typically fall into a coma, where their eyes are closed, they can't be roused, and they lack awareness. Some then enter PVS, a state of unawareness in which they cycle between sleep (eyes closed) and wakefulness (eyes open). MCS is a transitional state into which patients move from a coma or PVS and from which they might awaken, says Joseph T. Giacino, associate director of neuropsychology at JFK Johnson Rehabilitation Institute in Edison, NJ. MCS is characterized by limited awareness and inconsistent but clear evidence of consciousness.³ The United States is estimated to have hundreds of thousands of MCS cases and tens of thousands of PVS cases.

PET studies of vegetative patients indicate that the primary sensory cortices respond to pain and sounds, but that higher-order associative cortices do not. For minimally and fully conscious people, in contrast, sounds activate associative areas as well.⁴ The first published fMRI investigation of minimally conscious patients appeared this past February. (Laureys had already published a PET study of a single MCS patient.⁵) Joy Hirsch, of Columbia University Medical Center, and Nicholas D. Schiff, of Weill Cornell Medical Center, headed a team that exposed two MCS patients to recorded narratives. The stories elicited similar brain activity in the patients and in healthy controls. But when played in reverse, the stories strongly activated only the controls' cortices.⁶

© 2005 Taylor & Francis Group

Sentences were read to a patient who was in a vegetative state 13 months after he suffered a stroke. Functional magnetic resonance imaging revealed bilateral activation in his superior temporal lobes (top) similar to the activation found in healthy controls (bottom). The patient later emerged into a minimally conscious state. (From A.M. Owen et al., Neuropsychol Rehab, 15:290–306, 2005.)

To account for the latter finding, Hirsch suggests that only fully conscious brains "seem to be very engaged by ambiguous stimuli," such as a nonsensical reverse recording. The first finding surprised her team and has since been replicated, she says. It raises the question: Why weren't the patients fully conscious, given that their brain activity was so similar to that of normal subjects? The researchers theorize that the patients' low cerebral metabolism did not reach a level needed to spontaneously drive the neural networks necessary for cognition and interaction with the environment.

The Schiavo controversy began to heat up several weeks after this study appeared in the journal Neurology. "My mailbox filled up almost immediately," recalls Hirsch. Some correspondents, she says, related how they had long cared for minimally conscious relatives even though doctors had repeatedly stressed that the patients "could not appreciate or were not aware" of the family's presence. These correspondents wrote that, because of Hirsch's paper, they knew now that their attentiveness "had meant something" to their loved ones. Other E-mails sought help. A typical message, according to Hirsch, was: "My brother has been comatose for five years. Will you please see him?"

PAIN, DESIRE, AND SLEEP

Most patients with consciousness disorders never, in fact, undergo fMRI or PET scans. Whether brain imaging becomes more commonplace could depend on the outcome of a project that Giacino began in 2002. In collaboration with Schiff and Hirsch, he is observing whether fMRI scans of 15 MCS patients correlate with their bedside behavior and predict their chances of recovery.

Laureys, meanwhile, continues to examine unconscious subjects' responses to pain. He says that he halted an imaging study "earlier than planned because the results were so convincing that these [MCS] patients did feel pain," even though their behavioral deficits suggested otherwise. Using conscious volunteers, his lab has recently developed an fMRI-based method to infer a person's pain level without requiring a subjective report.

© 2002 Elsevier Science

One minute of electrical stimulation was delivered to the wrists of 15 healthy controls and 15 patients in a persistent vegetative state (PVS). Positron emission tomography indicated the activation of various brain regions (red), including the primary somatosensory cortex and thalamus. Blue areas, which were activated less in patients than in controls, included higher-order cortices. (From S.L. Laureys et al., NeuroImage, 17:732–41, 2002.)

Owen and Laureys are searching together for evidence that unconscious patients can have desires or intentions. Owen notes that when people imagine squeezing their hands, the motor cortex is activated. "What we're doing in healthy volunteers at the moment is looking at the extent to which it's possible to reliably detect people thinking certain things," he says. That effort was "fantastically successful in controls, and we are waiting to try it in the first patient that comes along."

Laureys is collaborating with other investigators as well. A project with Dehaene and Lionel Naccache, a neurologist at the Hôpital de la Pitié-Salpêtrière in Paris, will expose unconscious patients to auditory stimulation. The team will track brain activity by combining fMRI with event-related potentials, measurements related to electroencephalograms (EEGs). Dehaene predicts that the stimulus will be processed in early cortical areas but will not ignite the global brain pattern that "constitutes the substrate of a conscious state."

Giulio Tononi, a psychiatry professor at the University of Wisconsin-Madison, plans to examine Laureys' patients by combining transcranial magnetic stimulation with EEGs. In preliminary experiments, his lab has used this approach to infer that cortical connectivity drops dramatically when humans fall asleep. At issue is whether a similar drop occurs in minimally conscious and vegetative people. Laureys "has these patients who are extremely difficult to enlist, and he has really been a pioneer in studying them carefully in a scientific manner," says Tononi. "We have pioneered this new technique. So we need to put the two things together. It clearly can be done."

Consciousness Theories

© 2004 Tononi

According to the information integration theory, the amount of consciousness available to a system – even a machine – can be measured as a value: Φ. A heterogeneous arrangement of connections and an integration of elements both increase Φ (top). Lower Φ values result when connections are less heterogeneous (middle) or elements are less integrated (bottom). (From G. Tononi, BMC Neurosci, 5:42, 2004.)

Conferences about consciousness are "getting better in the sense that the approach is more often reasonably scientific and is less philosophical or New Age," observes Giulio Tononi, a psychiatry professor at the University of Wisconsin-Madison. But meetings can bog down in "enormous disagreements" over whether animals, machines, and vegetative patients are conscious and "That highlights the need for a theory," he says.

Tononi asserts that a theory must explain why consciousness requires the brain's thalamocortical system but not the cerebellum, and why it fades during dreamless, slow-wave sleep. He recently outlined his information integration theory of consciousness.¹ It mathematizes phenomenology – subjective experience – and creates a value, Φ, that indicates a system's ability to integrate information. Tononi posits that the presence of this ability distinguishes conscious states from unconscious states.

The theory is "way too general," contends Stanislas Dehaene, who performs cognitive neuroimaging experiments in Orsay, France. "It does not lead, for the moment at least, to specific predictions" for such experiments. Another objection is that the theory allows for machines to be conscious if their circuits have a high enough Φ value. Christof Koch, a consciousness expert at the California Institute of Technology, points to a third problem: the difficulty of computing Φ for "networks of more than 10 or 20 neurons." Tononi says that he has not yet used "realistic neural models" to test his theory, though he has created such models to study related questions.

The global neuronal workspace hypothesis is another current effort to make systematic sense of consciousness. Dehaene and two other French scientists, Jean-Pierre Changeux and Lionel Naccache, have recently championed this hypothesis, which was originally developed by Bernard J. Baars, of the Neurosciences Institute in San Diego. It proposes that consciousness occurs when neurons with long-distance connections spontaneously broadcast signals throughout the brain. This sudden activity makes information available to various perceptual and cognitive processes.

In our model, "there is spontaneous activity all the time," explains Dehaene. This crucial property "doesn't have to wait for external inputs," and it depends on ascending inputs to the cortex and thalamus. In many consciousness-impaired patients, Dehaene adds, "I think it's the source of the ascending input from the brainstem which has deteriorated.... In other cases, it can be that the network of interrelated cortical areas has deteriorated."

Tononi is critical of the workspace hypothesis, characterizing it as more a metaphor than a theory, and arguing that its notion of information is ill-defined. He also asserts that at least one widely broadcast signal is not crucial for generating consciousness. Removal of the locus ceruleus, a nucleus whose neurons release noradrenaline throughout the brain, does not appear to damage consciousness, he says.

"An information integration theory of consciousness," Tononi G, BMC Neurosci , 2004 Vol 5, 42