Consciousness in questions

Abstract

What is it like to be conscious? Whereas nobody knows, by definition, what is it like to be unconscious, everybody knows what is it like to be conscious. Those of our mental states that are conscious have a distinctive phenomenal character, which we can experience and report. For example, there is something about seeing things that has its own distinctive character, which differs from hearing or smelling them. This property of having a distinctive phenomenal character is shared, not only by conscious mental states built up from sensory ingredients but also extends to states arising from ‘within’—such as beliefs, thoughts, desires, hopes or intentions, all of which correspond to distinctive experiences. What gives conscious mental states their distinctive quality is their content: they are about something or, better, they represent something; and this representation determines their phenomenal character. This representationalist view raises a key question for the scientific study of consciousness, particularly for the study of neural correlates of conscious experiences: do the neural correlates of conscious experience relate to the content of that experience? According to Chalmers (1998), a neural representational system counts as the correlate of an experience if the neural correlates are themselves sufficient for that experience. So, ‘if there are neural representational systems whose contents match the contents of experience in this way, then knowing what the contents of an experience are helps us identify its neural correlate. Conversely, if one already knew what the neural correlate of an experience was, its role in information processing could tell us what contents the experience has’ (Siegel, in Bayne et al., 2009, p. 192). Controversial as this statement may be (there are indeed non-representationalist views of consciousness as well, see Hellie, in Bayne, 2009, pp. 563–6), it sets the stage for studying the neural constituents of consciousness. Correlations between states of consciousness and brain activity have been looked for experimentally, with the aim of determining the minimal changes in neural representation that would suffice for conscious content to appear or disappear. Yet things are not that simple. Consider for example, the effect of experiments involving transcranial magnetic stimulation of the visual cortex in healthy subjects: a stimulation pulse applied to the lateral occipital zone (corresponding to cortical area V5, a region containing many motion-sensitive neurons) results in the subject consciously perceiving motion within the visual field. However, this conscious content cannot be the mere effect of activating movement sensitive neurons because a second pulse applied 15–45 ms later on cortical area V1 abolishes the effect (Pascual-Leone and Walsh, 2001). This finding speaks against the ‘microconsciousnes’ view of visual awareness, according to which a conscious experience would be elicited by the activity of a single specialized neuronal population. Instead, it indicates that the level of activity in V1 gates access to the visual experience of motion through a more complex cortical circuitry (Walsh, in Bayne et al., 2009, pp. 635–9). The effects of brain lesions extend this claim even further. Patients with visual neglect following parietal damage lack phenomenal experience of objects in one half of the visual field; yet presentation of visual stimuli in the neglected hemifield evokes activity in the primary visual cortex, suggesting that the key region for eliciting visual experience might not be located in area V1, but in areas remote from it, possibly in the parietal cortex (Rees, in Bayne et al., 2009, p. 210). The conditions for drawing inferences between neural correlates and conscious content are not easily met because even if a good match is found between the content of the neural representation and the content represented in consciousness, the significance of this match remains to be interpreted. ‘If it is interpreted in terms of “identity”, then all the experiential properties must in principle be found in the neural representational system’ (Hohwy, in Bayne et al., 2009, pp. 203–7). However, many philosophers resist this idea, arguing that the conscious experience has representational properties that the neural content does not (and cannot) have. One such property is perceptual coherence of experience: we are experiencing things as coherent wholes, not as assemblages of attributes, whereas the nervous system is supposed to operate on a building block principle by processing each attribute with specialized processors. The same can be said if one considers conscious perception at the macroscopic, cognitive level: although psychophysical experiments demonstrate that the different properties of an object are processed separately at the input stage, object representations become holistic in more central systems dealing with cognitive activities such as selective attention, decision making and declarative memory. How can the distributed nature of neural processing be reconciled with the intrinsically unified nature of conscious experience? Consider the case of perceiving an external object visually. Neurophysiologists have proposed the existence of a process of integration that would bind together the features of that object (colour, shape, texture, etc.) to form a unique percept. This process is materialized by the occurrence, in areas of the visual system involved in processing the object, of fast oscillating potentials (in the gamma frequency range, ∼40 Hz). It is thought that these oscillations reflect the temporal correlation of the neuronal populations selected by the stimulus; accordingly, they would play a role in synchronizing the activity of these distributed populations into a functionally coherent cell assembly. Crick and Koch (1990), elaborating on this hypothesis, suggested that only appropriately bound neural activity could enter short-term memory and hence become available for access to phenomenal awareness (Engel, in Bayne et al., 2009, p. 322). Thus, in theory, the temporal patterning of neural activity would be critical for the emergence of conscious states (Engel and Singer, 2001). This theory, although it clearly accounts mechanistically for the holistic features of neural activity that represent an external stimulus, still remains far from explaining why this holistic activity should correlate with conscious experience. Indeed, temporal correlation with gamma oscillations has been found in anaesthetized, unconscious animals, where no phenomenal unified experience about the stimulus can be expected. The existence of cognitive binding for object properties also does not account for how these holistic representations become accompanied by subjective conscious experience. As a matter of fact, cognitive binding is a feature common to human beings and information processing devices such as robots or machine vision systems, for which the discussion about subjective experience becomes largely irrelevant (Revonsuo, in Bayne et al., 2009, p. 102). Thus, neural binding and its cognitive counterpart are still removed from the phenomenal binding experienced in conscious states. Even if one (tentatively) accepts the validity of the binding principle as a factor of conscious awareness, one would still have to understand how locally unified contents can themselves be bound together to form a unified phenomenal field. The unity of phenomenal consciousness is indeed the ‘hard problem’ of consciousness (Alter, in Bayne et al., 2009, pp. 340–3): our sensory experiences, as well as our internal thoughts and images, intentions, motivations and emotions are integrated in such a way that, when conscious, we experience ‘being one single person within one single world’ (Metzinger, 2003). But how it is that a person’s simultaneous experiences are usually co-conscious is a matter of controversy. Atomistic or ‘building block’ approaches to consciousness (e.g. Searle, 2000) tend to account for only particular conscious states—a pain, a visual experience, a conscious thought–on a case-by-case basis, rather than accounting for the subject’s entire phenomenal field at once: they posit the coexistence of one mechanism responsible for making mental states conscious and another for making them co-conscious. On the contrary, holistic theorists invoke a single mechanism for both making mental states conscious and putting them together into a unified phenomenal field (Bayne and Levy, 2006). Accordingly, consciousness would arise as a single global state, of which the various conscious states—bodily sensations, thoughts, perceptual states, affective states, etc.—would be components or abstractions (Bayne, in Bayne et al., 2009, p. 648). The current neural models available for phenomenal binding tend to reconcile the two approaches. One such model is the so-called global workspace theory. According to its proponent (Baars, 1989), this theory ‘was derived from applications in artificial intelligence, as a computational tool to enable multiple specialized processors to cooperatively solve problems that they could not solve individually’ (Baars, in Bayne et al., 2009, p. 332). The global workspace is seen as a sort of central memory store that can be accessed by many input assemblies of neurons which, when active, determine the specific content of consciousness. These input assemblies form large receiving collections, which themselves assemble to form larger contexts which can select, evoke and shape the contents of the global workspace. When a given coalition of contexts is stabilized, conscious experience may arise in the interaction between the receiving assemblies and the content of the global workspace. This rather abstract architecture has been refined and implemented in a more likely and testable neuronal model, based on well-identified features of brain networks. This ‘neuronal’ global workspace model (Dehaene et al., 1998) postulates that parallel, non-conscious, modular processors compete for access to the global workspace. The global workspace is seen as a capacity-limited system where modular processors exchange information. This information will form the content of subjective experience at any given time. Dehaene and his co-workers (1998) consider that the neural basis for the global workspace is the highly distributed set of cortical cells, the large pyramidal neurons with long axons mediating long-distance, cortico-cortical connections. These neurons, which interconnect distant specialized cortical and subcortical processors, break the modularity of the system. In addition, they exert a top-down modulation by conveying amplification signals to almost all cortical regions. Temporally amplified regions gain access to the global workspace, which determines the conscious content of the workspace. Empirical evidence for the global neural workspace model can be drawn from several sources. First, the putative anatomical support of the global workspace (the presence of large pyramidal cells with long axons and large dendritic trees) fits into the characteristics of brain areas such as the prefrontal, cingulate and parietal cortices, known to be tightly concerned with conscious processing. Secondly, cognitive neuroscience experiments using neuroimaging have revealed patterns of cortical activation compatible with the theory. These experiments show that the conscious reportability of perceptual stimuli is associated with both an increased activation of areas coding for the relevant content and an activation of the prefrontal–parietal areas: in other terms, consciousness is associated with the amplification of the modular processor, combined with the activation of the long-distance interconnections forming the global neuronal workspace. This same prefrontal–parietal network is spontaneously active in waking subjects at rest, and its activity decreases in states where consciousness tends to vanish (e.g. certain states of sleep). In contrast, a high level of activity is restored during rapid eye movement sleep, which is consistent with the conscious dreaming activity that occurs during this state of sleep (Dehaene, in Bayne et al., 2009, p. 468). The unity of consciousness is a fragile condition, which may break down as a consequence of brain lesions, particularly those located in posterior parietal areas. A typical example is that of Balint’s syndrome following bilateral parietal lesions. Patients with Balint’s syndrome present with difficulties in localizing objects that they report seeing. They cannot reach for these objects and cannot fixate them by gaze. At the same time, they make errors in failing to bind the basic features (colour, size) of an object to the shape of that object; in some cases, features from other items in a display may even be incorrectly attached to the shape the patient sees, resulting in the experience of an illusory conjunction; for instance, in a display containing a blue ‘T’ and a green ‘X’, the patient may report perceiving a blue ‘X’ (Robertson and van Vleet, in Bayne et al., 2009, pp. 99–100). Thus, without an accurate spatial map of the external world, the phenomenal field of consciousness loses its internal structure. Other parietal lesions, more frequently located in the right hemisphere, can produce a severe disorganization of conscious experience related to the body, to actions or even to the self. Following such lesions, a patient may deny ownership of his/her paralyzed left arm and even misattribute it to someone else, in spite of contradicting evidence. Another condition is anosognosia, the denial of illness and the consequent false belief of being ‘normal’; anosognosic patients with hemiplegia following a right-sided lesion, when asked about their potential ability for performing actions, claim that they can perform any type of movement. Their false conviction of still being able to move persists even when sensory and visual evidence from the motionless affected limbs suggests that no movement has been performed. Anosognosia for hemiplegia is a paradigmatic example of a disorder of self-consciousness, which ‘is an impairment of awareness that does not affect the processing of all kinds of stimuli and events, but that is confined to the monitoring of a particular (motor) condition of the self’ (Berti, in Bayne et al., 2009, p. 54). This remark draws attention to a category of states of consciousness that do not relate to perceptual representations, but to representations that arise from within, in conjunction with the generation of actions. Action raises questions different from perception, because the phenomenal content specific to action consciousness refers to the author of that action. The question is no longer about ‘what is it?’ or ‘where is it?’, but ‘who did it?’: in other words, it is about the sense of agency, the ability to self-attribute one’s own actions. Research in this domain has focused on the way neural signals generated at the time of an action can provide cues for the sense of agency. The current conception is that action-related signals (first person signals originating from within the agent’s body and third person signals originating from outside) are dynamically compared with an internal ‘forward model’ generated prior to the action and anticipating its effects (Wolpert et al., 1995). Accordingly, a satisfactory match of action-related signals with anticipation should identify a self-generated action and therefore fulfil the conditions for sense of agency. The paradox, however, is that a vast majority of our actions are performed automatically, without a conscious phenomenal experience. This is a constraint of goal-directed movements, which have to remain non-conscious for the sake of rapidity and accuracy. Conversely, it is common experience that movements executed under conscious control (e.g. during the first attempts at learning a new skill) are usually slow and inaccurate. Due to these constraints, action awareness appears to be limited to the period of time that precedes the action, when the anticipation is built, whereas the subject remains unaware of the execution itself. It is only in the case of a mismatch that, for example when the action fails, the subject may become aware of what he/she was doing (Blakemore, in Bayne et al., 2009, p. 11). These considerations lead to the curious conclusion that action consciousness and the conscious sense of agency arise as a consequence of action failure! (Jeannerod, 2003). The slowness of access to consciousness, as revealed by action studies, has deep theoretical implications. Whereas the fact that awareness of action is a mainly post hoc phenomenon seems to exclude the possibility for a causal role of consciousness, we still perceive ourselves as causal, a dissociation which has been considered by some as the origin of an ‘illusion’ (e.g. Wegner, 2002). According to these authors, the sense of free will is a cognitive phenomenon based on the (false) belief that our thoughts can have a direct influence on our behaviour. This belief would be reinforced by the fact that conscious thought and the connected action usually appear as a phenomenal sequence of events; from this observation, the self would tend to build a cause-and-effect story. However, regularities and temporal contiguity cannot substitute for causation. In addition, there are experimental situations, including those used by Wegner (2002), where a subject can easily be misled into thinking that he/she is exercising control to produce an outcome that is in fact produced by someone else (see Pink, in Bayne et al., 2009, pp. 299–300). The idea of a causal role for consciousness in action generation actually relies on a misconception of the respective roles of neural and mental activity in the determination of behaviour. Rather, conscious will should be considered as a consequence, not a cause, of the activity of a neural system, distinct from that responsible for the production of the action. Its function would be to bridge the gap between cognitive states prior to the action, such as intending, planning and the cognitive states based on knowledge of the action’s results after it has been performed. The confrontation of the two states would provide cues for the phenomenal experience of the self’s continuity over time, as well as for the feeling of being author of the action. This mechanism breaks down in pathological conditions like schizophrenia. As proposed by Coltheart (2007), delusions suffered by schizophrenic patients, such as the sustained impression of being controlled by external forces and having lost conscious will, might rely on an impossibility of connecting their action-related cognitive states correctly with each other, with the consequence of misattributing their own intentions and actions to alien forces. All the above questions, together with many others, are raised in The Oxford Companion to Consciousness, co-edited by Tim Bayne, Axel Cleremans and Patrick Wilken. This 670-page book covers the different facets of the phenomenon of consciousness in articles written by over 250 experts in philosophy, neuroscience, psychology and computer science. The most burning issues are treated by several authors of different disciplines in separate articles, providing the interested reader with a truly critical and interdisciplinary overview of the problem. Cross-references between entries are clearly indicated. Generous lists of are The of the articles to and see for example, the on consciousness in or that the problem’ do that the is when the is if it is each time the to This is an book for and of consciousness at all