7 July 2010. This is an English version of an article by Jennifer Altman in Alzheimer Actualités, a newsletter published in French by the Ipsen Foundation. The Alzforum editors gratefully acknowledge the Foundation’s generosity in making this summary freely available.
Consciousness is like beauty: We all know subjectively what it is, but it is very hard to define concisely. Consequently, the study of consciousness has until very recently been mainly the concern of philosophers. The development of sophisticated imaging technologies and high-powered computer analysis in the past 25 years has finally provided tools for accessing the neural correlates of the conscious state. Theoretical models of consciousness are being tested, our understanding of the role that consciousness plays in our cognitive processes is being refined, and methods are being devised for examining the residual consciousness of brain-damaged patients. All this and more was discussed at the Neuroscience Colloquium held in Paris on 3 May 2010, organized by Stanislas Dehaene (Collège de France and Unité INSERM-CEA de Neuro-Imagerie Cognitive, Gif sur Yvette, France) and Yves Christen (Fondation IPSEN, Paris, France).
For scientists, definitions are essential. It says much about the difficulty of defining consciousness that many of the speakers started their talks with their own definition. A minimal description is that the subject is awake, aware, and able to communicate about his/her internal state with others. Assessing another’s state of consciousness is more problematic, especially if communication is not possible. Because of the subjectivity on which these descriptions depend, consciousness has until recently not been considered a topic accessible to scientific investigation, but powerful methods for imaging brain activity are now providing researchers with objective ways to look into the brains of others and measure their responses to both external stimuli and changes in internal state. Not only the state of consciousness, but also some of the unspoken contents of the mind can now be examined.
The toolkit for probing consciousness now includes functional brain imaging, multi-channel recording of the brain’s electro-encephalographic activity (EEG), magneto-encephalography recordings, transcranial magnetic stimulation, and electrophysiological investigations in brain surgery patients. As the techniques have become more powerful, so, too, the ways of testing subjects have become more sophisticated, so that mental tasks, complex planning, and even purely introspective states are being investigated.
Although recording the electro-encephalogram (EEG) with electrodes on the scalp has been used for many years, its application has been vastly refined by computer analysis of complex signals, which enable the identification of fluctuations evoked by specific events and the location of their source within the brain. Imaging techniques have also brought huge advances in neurosurgery, allowing the placement of fine electrical probes in the brains of awake patients undergoing treatment for brain tumors or epilepsy. The probes are primarily used by the surgeons to guide the operation but are also being used, with patients’ permission, for research purposes. Information gathered through these various windows is slowly being pieced together to build up a picture of the operations of what is often termed the mind—the complex of mental abilities that we believe characterizes human beings.
The cerebral cortex is essentially a huge information processor that connects the body with the world around it. Each local area of the cortex deals with certain aspects of the incoming signals or with producing responses to them.
The concept that these local areas are composed of networks of neurons that “talk” to networks in neighboring areas is well established, but the problem remains of how the activity in all the areas is coordinated to produce a seamless response.
One hypothesis, the global neuronal workspace, envisages that the local networks form processing “hubs” linked together, and especially to the prefrontal cortex and the central areas of the thalamus, to form a meta-network (Jean-Pierre Changeux, Collège de France et Institut Pasteur, Paris, France; Dehaene). The substrate for this is the long axons of the cortical pyramidal neurons that form the white matter tracts running in both directions between the higher association areas and linking them to perceptual cortical areas (Changeux). Although these tracts exist in other mammals, their volume is many times greater in humans, a far bigger increase than that of the grey matter of the neocortex. The density of the dendrites of the pyramidal neurons, which receive inputs from the long-distance connections, increases in parallel throughout the mammals, as does the volume of the prefrontal areas of the cortex.
According to the global neuronal workspace hypothesis, information becomes conscious when it is made available to the whole cortex through reverberating electrical activity in the long-distance connections, particularly when the prefrontal cortex is involved (Dehaene). This global activation, termed ignition, depends on the excitatory glutamate receptors on the synapses used by the feedback connections: Known as NMDA-type receptors, they require a certain strength of input before they reach the threshold to transmit a signal. Ignition, an all-or-none event, occurs only when the bidirectional connections are sufficiently excited to exceed this threshold (Changeux). Experiments using masked stimuli reveal that until ignition happens, perception is below the level of awareness; once triggered, subjective reports can be made. The key role of the prefrontal lobes and white matter tracts in conscious awareness is highlighted in conditions such as schizophrenia and multiple sclerosis, where impairment of ignition and subjective reporting correlates with damage to these structures (Dehaene).
One limitation of the global workspace model is that it considers only the cortico-thalamic connections. A fuller account must include the awareness of internal state that gives rise to the sense of self (Antonio Damasio, Brain and Creativity Institute, University of Southern California, Los Angeles). In this view, normal consciousness is the integration of the processes that give rise to the sense of self into an awake mind; there is no clear separation between body and brain. The self has three stages: the protoself, core self, and autobiographical self. The protoself, whose main role is homeostatic, produces the spontaneous feelings of the living body. It is embedded in a network of centers in the brain stem, mid-brain, hypothalamus, and the deeper, older areas of the cortex, such as the insula and anterior cingulate. It also encompasses neurons in the nuclei of the reticular formation, which branch throughout the cortex, releasing the neuromodulatory transmitters noradrenalin and serotonin; activation of these pathways affects which incoming information is attended to. Whenever the organism interacts with an external object, the protoself becomes modified, generating a pulse in the core self. When pulses in the core self generated by memories are activated in a coherent pattern, they give rise to the autobiographical self.
The thalamus, a mid-brain structure through which most of the inputs to and outputs from the cortex are connected, has a prominent role in both these models—which is significant because the thalamus also regulates the sleep-wake cycle. In particular, the matrix and intralaminar nuclei, which may be important for decisions to engage with stimuli (Michael N Shadlen, University of Washington, Seattle), are implicated in patients recovering after long-lasting loss of consciousness, either through their reconnection with the cortex (Steven Laureys, Université de Liège, Liège, Belgique) or as a result of electrical stimulation (data from N. Schiff’s lab, reported by Laureys and Shadlen).
Many experimental paradigms, by necessity, examine temporary changes in conscious content activated by conjunctions of neural events triggered by responses to external stimuli (Dehaene; Lionel Naccache, Institut du Cerveau et de la Moelle épinière, CHU Pitié Salpêtrière, Paris, France; Damasio). However, much of the brain’s activity seems to be independent of environmental input—intrinsic activity that is usually discarded in the analytical procedures that generate the familiar brain-scan maps but also represented by very slow waves in the EEG (Marcus Raichle, Washington University School of Medicine, Saint Louis). This resting activity, or default network, which seems to be responsible for most of the brain’s huge energy consumption, changes independently of, or sometimes in anticipation of, external stimuli rather than in response to them. It may have a priming or alerting function, providing a temporal coherence between the activities in different processing hubs.
How coherence is achieved between related pieces of information—the so-called binding problem—is still unresolved. Although we know that the barking dog is standing on green grass, the brain processes the sensory input in fragments in many separate channels: sound, color, shape, contrast, and so on. The various attributes converge again at later stages in the processing networks, but what ensures the bark is attached to the dog and not to the grass? One proposal is that signals in individual neurons coding for the mental representation of the same object synchronize by using a frequency component of the EEG as a carrier wave (Pascal Fries, Ernst Strüngmann Institute, Frankfurt, Germany). These signals will be encoded by one phase of the wave and thus be reinforced, whereas those from unrelated objects will be out of phase and so cancel out. Simultaneous recordings from several areas in the monkey brain support previous evidence that the signals in the forward connections ride on the gamma-frequency oscillations. The feedback signals related to the same object are probably distinguished from the forward signals by using a lower-frequency carrier wave.
A fundamental aspect of animal behavior is making decisions about what actions to perform. Studying the neural basis of decision-making may provide insights into how the brain reasons: Deciding to do may have given rise to deciding to consider or to engage (Shadlen). Recording from the monkey cortex while the animal performs a visuo-spatial decision task is revealing how individual neurons respond during different stages of the process. In the higher centers of the visual cortex, some neurons with spatially selective responses fire persistently while the monkey is deciding to perform its chosen movement—these signals provide a short-term information store, allowing the monkey to accumulate evidence in favor of the impending decision; in other words they give freedom from immediate response. In the lateral inferior parietal cortex, a visuo-motor association region the neuron signals indicate that inferences have a temporal component—they are made on data streams rather than data sets—and that the strength of the neuronal response combines evidence, expected reward, and cost with elapsed time, ultimately providing both a decision and an estimate of the confidence associated with that decision.
Changing Our Minds
Clearly, a great deal of the neural processing in local areas happens outside awareness. Much more processing is apparent from fMRI scans during choice experiments than the subject reports (John-Dylan Haynes, Charité-Universitätsmedizin, Berlin, Germany). The decision-making process involves a cascade of unconscious steps, first in the frontal regions of the cortex, then in the pre-motor areas, and lastly in the motor areas, where the movement is generated. Even in the higher executive regions, content-specific material does not necessarily reach awareness, and activity in the medial prefrontal and parietal cortices reveals conscious but concealed plans as well as information that has not yet resulted in a conscious decision.
One surprise is that we seem not to be aware of the experience of moving but only of the intention to move (Angela Sirigu, Institut des Sciences Cognitives, Bron, France). Awake patients undergoing brain surgery were asked to report on their sensations while various motor areas of the brain were electrically stimulated. When the inferior parietal lobe, a sensorimotor association area, was stimulated, patients reported the desire and intention to move and produced illusory movements, but no movements were triggered. In contrast, they were not aware of the movements that actually resulted from stimulation in pre-motor areas.
Since the 1950s, a lot of effort has gone into trying to influence subconscious processes, mostly without success. Yet recent experiments are showing that processing can be modified by subliminal stimuli (Mathias Pessiglione, Inserm U610, CHU Pitié Salpêtrière, Paris, France; Moran Cerf, California Institute of Technology, University of California, Los Angeles). Motivation involves the subcortical motor planning areas in the basal ganglia as well as the cortex: patients with bilateral lesions in the basal ganglia lose all spontaneous behavior, although they can move in response to instructions (Pessiglione). Behavioral testing demonstrates that subliminal cues can be learned and do motivate behavior when they are rewarded, although the effects are small. The effectiveness of the cues seems to be governed by neurons releasing dopamine in the basal ganglia. These tests are aiding understanding of different types of patients, some of whose motivation is low, as in depression, and others who are over-motivated and impulsive.
Questions such as whether our internal mental state or the external environment has the stronger influence are being investigated by recording with electrodes implanted into the brains of patients in surgery for epilepsy. The responses from single neurons in areas concerned with visual processing, memory, emotions, and attention are registered as the subject views several favorite images (Cerf). The neuronal responses are highly selective—for example, a neuron in one patient fired repeatedly when a picture of Halle Berry was displayed, and even when the patient imagined the picture. An exciting recent experiment shows that neuron firing can be influenced rapidly, strongly, and reliably in thought experiments. When merging a preferred and a distracting image, whose contrast levels are determined by the strength of the neuronal response, the patient can learn to control the firing by imagining the preferred image and making it “win” over the distracting picture presented simultaneously on the screen.
Communicating our internal state requires the use of language, but how language evolved is still hotly debated (Herbert S Terrace, Columbia University, New York). Years of painstaking experiments failed to teach apes to converse using symbols as substitutes for words, demonstrating that learning theory alone cannot account for the development of language, as claimed by behaviorist psychologists. Chomsky’s argument that language must have emerged fully fledged to serve thought runs contrary to the principles of Darwinian evolution. It seems more likely that language is based on a pre-existing cognitive capability, most likely with a social function such as child care. The development of pre-verbal communication between infant and caregiver is rooted in eye contact and direction of gaze. This could have emerged only after early humans became bipedal, with the consequences that babies were born less well developed and more dependent on their parents and that they were carried in the parents’ arms rather than on the back or belly, affording direct eye contact.
Communication as a measure of consciousness becomes crucial in brain- damaged patients unable to move (Naccache; Laureys; Adrian Owen, University of Cambridge, Cambridge, UK). Four main categories of patients have been identified: comatose patients have preserved non-conscious processing but impaired global cortical activity and its activation through the ascending reticular system; patients in the vegetative state have activity in disconnected patches of cortex and a functioning reticular system but no intense global cortical and thalamic activation; minimally conscious patients have impaired, fluctuating cortical activation; and locked-in patients are conscious but paralyzed (Naccache; Laureys). Activity in the precuneus area of the cortex, which seems to be essential for global activation, and in the resting state are also a good indicators of the state of consciousness (Laureys).
Although fMRI imaging is revealing cortical activity in many of these patients, the big problem is disentangling conscious from non-conscious processing. In an “oddball” test, the patient is asked to identify a sequence of five identical chords from several sequences in which the fifth note is different. Local irregularities elicit only local, non-conscious processing; global activation is seen only in patients conscious of the global regularity of the repetitions (Naccache). A version of the test for use with EEG recordings at the bedside is being developed.
The power of the imagination is being harnessed to develop a way of communicating with such patients (Laureys, Owen). When normal subjects are asked to imagine playing tennis, the supplementary motor area lights up in fMRI scans; when asked to imagine navigating between rooms in their houses, the parahippocampal area is activated. Patients unable to move manage to communicate with simple yes-no answers by imagining playing tennis when the answer to a question is yes, and navigating for no. So far, only simple questions have been used, but the technique could be developed to interrogate patients about how they feel and what they are aware of.
The use of fMRI methods for assessing consciousness in those unable to communicate is fraught with difficulties of interpretation, but careful testing is indicating that about 40 percent of those initially diagnosed as vegetative by clinical examination might be in the minimally conscious state. These advances also raise substantial ethical questions, as well as requiring considerable management of expectations and public interest. However, by providing tools for more accurate determination of patients’ state, they enable better treatment, especially in the treatment of pain (Laureys, Owen).
Looking to the Future
This is a very young field, and already, to apply the metaphor of the global workspace hypothesis, it has reached ignition. So far, some of the common, basic mechanisms of consciousness are being described. Ahead lies the long road to understanding what makes the consciousness of each individual unique and identifying the processes that make us the creative individuals that we are.—Jennifer Altman.
Jennifer Altman is a science writer in Todmorden, U.K.