Consciousness: Theoretical Approaches

Consciousness: Theoretical Approaches 2 Tim Bayne and Jakob Hohwy Abstract This chapter reviews some of the central theoretical challenges confronting the search for the brain basis of consciousness and develops a conceptual framework for tackling these challenges. At the heart of the search for the neural basis of consciousness is the notion of a neural correlate of consciousness. Identifying the neural correlates of consciousness requires that we acknowledge the various aspects of consciousness, for each of the aspects of consciousness raises its own set of methodological challenges. We examine the question of whether an account of the neural correlates of consciousness can be used to ascribe consciousness to creatures that lack the capacity to report their experiences, and we ask whether it is possible to go beyond the neural correlates of consciousness by providing neurally-based explanations of consciousness. 2.1 Introduction In this chapter, we review three aims of the science of consciousness for which neuroimaging is potentially relevant. The first aim is that of identifying the neural correlates of consciousness (NCCs), where a neural correlate of consciousness is a neural state that suffices for a certain kind of conscious state. The second aim is that T. Bayne (*) Department of Philosophy, School of Social Sciences, The University of Manchester, Oxford Road, Manchester M13 9PL, UK e-mail: [email protected] J. Hohwy Department of Philosophy, Monash University, Clayton, VIC 3800, Australia e-mail: [email protected] A.E. Cavanna et al. (eds.), Neuroimaging of Consciousness, 23 DOI 10.1007/978-3-642-37580-4_2, © Springer-Verlag Berlin Heidelberg 2013 24 T. Bayne and J. Hohwy of developing robust tools for the ascription of consciousness. The hope is that once we have identified the neural correlates of consciousness we can use our knowledge of them to determine whether or not a target creature is in a particular kind of con- scious state. A third aim for the science of consciousness is to go beyond the identification of the neural correlates of consciousness by providing mechanistic explanations of consciousness. We will consider how neuroimaging might bear on each of these three aims, but first we present a framework for thinking about con- sciousness that informs our discussion of these issues. 2.2 A Conceptual Framework for Consciousness A central distinction is that between specific conscious states and the generic state of being conscious (what is sometimes called ‘creature consciousness’). Specific conscious states are individuated in terms of their content. For example, we can distinguish the state of smelling a rose from the state of smelling coffee, and we distinguish each of these states from the state of seeing a face. We can say that a creature is in the generic state of consciousness if, and only if, it is in some con- scious state or another. For the most part the scientific study of consciousness has focused on the study of specific conscious states. For example, a typical paradigm for investigating con- sciousness employs binocular rivalry, where a conscious monkey or human is shown different stimuli to the two eyes and the conscious content of their experience changes between what is shown to the two eyes. The aim, using various neuroimaging tech- niques, is to identify the changes in neural activity that correlate with the changes in conscious experience. Here, investigators are not interested in the domain-general mechanisms that underlie consciousness as such, but are instead interested in the particular mechanisms that are implicated in having certain contents in conscious- ness (namely, those that characterise the content of binocular rivalry). Most discussions of specific conscious states focus on phenomenal conscious- ness. A mental state is phenomenally conscious if there is ‘something it is like’ for the creature in question to be in that state (Nagel 1974; Chalmers 1996). Phenomenal consciousness is most closely associated with various forms of sensory experiences, such as perceptual states, bodily sensations, and affective states. However, not all conscious states are sensory, for consciousness also includes thoughts of various kinds—beliefs, desires, and intentions. It is controversial whether conscious beliefs, desires, and intentions are forms of phenomenal consciousness or whether—as many theorists believe—such states are conscious only in some non-phenomenal sense (Bayne and Montague 2011). At any rate, a full account of consciousness must accommodate all kinds of specific conscious states, both sensory and non-sensory. Although the science of consciousness has tended to focus on specific conscious states, in recent years there has been increased interest in the mechanisms that underpin the generic state of being conscious. Some research groups have employed neuroimaging to explore the contrast between the presence and absence of con- sciousness in the persistent vegetative state (Laureys 2005; Bekinschtein et al. 2009; 2 Consciousness: Theoretical Approaches 25 Boly et al. 2011), whereas others have used neuroimaging to examine the global changes to the conscious states that are seen in anaesthesia (Alkire et al. 2008). In addition to specific conscious states (the ‘contents of consciousness’) and the generic state of being conscious, we need also to recognise a third aspect of con- sciousness: what we will call modes of consciousness. We can think of modes as non-specific states of consciousness. Distinct modes of consciousness are associated with normal wakefulness, REM dreaming, hypnosis, mild anaesthesia, epileptic absence seizures, and the states associated with various kinds of consciousness- altering drugs such as LSD. Modes of consciousness are not defined in terms of their content in the way that specific conscious states are, but are instead characterised by appeal to the subject’s cognitive and behavioural capacities (or the lack thereof). For example, in some modes of consciousness the subject might have introspective access to and atten- tional control over their specific conscious states whereas in other modes these capacities might be disrupted or even altogether absent. Modes of consciousness are often referred to as ‘levels of consciousness’ or ‘background states of conscious- ness’. We find both terms problematic: the former suggests that modes of conscious- ness involve different degrees of consciousness (which they may not do), and the latter suggests that modes are recessive features of the field of consciousness against which the contents of consciousness stand out (which they aren’t). Although we have distinguished these three aspects of consciousness (specific states, modes, and the generic state of being conscious), we do not mean to suggest that these three aspects of consciousness are independent of each other. On the con- trary, it is clear that there are deep and important relations between these three aspects of consciousness and that a full account of consciousness will involve iden- tifying these relations. Indeed, one of the ways in which the neuroscience of con- sciousness might contribute to our understanding of consciousness is by advancing our understanding of these relations. We return to this point shortly. 2.3 The Neural Correlates of Consciousness We turn now to the notion of a neural correlate of consciousness (NCC). An NCC for a conscious state kind C1 is a neural state that is minimally sufficient for instances of C1 in the members of a certain population (Chalmers 2000; Hohwy 2007). The qualification of minimal sufficiency is included in order to ‘screen off’ those neural states that are of only causal relevance to the presence of C1. The overall state of an individual’s brain may suffice for one of its conscious states (C1), but such a state would not, intuitively, qualify as an NCC for C1 insofar as it would contain a great deal of neural activity that wasn’t directly required for C1. Note a couple of general features concerning the NCCs. Firstly, the term ‘corre- late’ is employed in order to leave open the precise relationship between neural states and conscious states. Although there is general agreement that the neural cor- relate of C1 is not a cause of or causal precondition for C1, there is debate about whether the relation between conscious states and their neural correlates is that of 26 T. Bayne and J. Hohwy identity, constitution, or realisation or whether—as some dualists hold—conscious states are merely correlated with neural states. Secondly, it is important to recognise than an NCC holds only relative to a cer- tain population. Some NCCs might hold with respect to a number of species, others might hold only with respect to a particular species, and still others might hold only with respect to the neurotypical members of a species. Indeed, we shouldn’t rule out the possibility that the neural basis of consciousness might be quite variable even across the neurotypical members of a single species and that the neural correlate of C1 in one person might differ in non-trivial ways from the neural correlate of C1 in another person. We return to this topic below. In light of the distinctions that we made in the previous section, we can see that there will be at least three different types of NCCs: there will be NCCs associated with specific conscious states of various kinds (such as smelling a rose), there will be NCCs associated with conscious modes of various kinds (such as normal waking awareness or dreaming), and there will be NCCs associated with the generic state of being conscious. Thus, there will be minimally sufficient neural correlates for dis- tinct aspects of consciousness, and each of these NCCs will raise its own set of issues. An important (but poorly understood) distinction that bears on specific NCCs concerns a contrast between two components of such correlates, what we might call their differentiating components and the non-differentiating components. The dif- ferentiating component of a specific NCC will be that component of its correlate that is selectively implicated in its content. For example, there is evidence that acti- vation of MT/V5 is implicated in conscious experiences of visual motion; thus, this might qualify as a differentiating correlate for the conscious state of seeing move- ment. But it is highly unlikely that MT/V5 itself suffices for experiences of visual motion, for one wouldn’t expect MT/V5 activation that was causally isolated from all other activity in the brain to generate experiences of visual motion. In order to function as a full correlate of experiences of visual motion, MT/V5 activity must be suitably integrated with certain kinds of non-specific activity, activity that is impli- cated in many kinds of conscious states. In other words, the full neural correlates of specific conscious states will have non-specific components that do not differentiate between the contents of those states. The distinction between differentiating and non-differentiating correlates has important methodological implications. The study of specific NCCs involves the use of individuals who are conscious throughout the procedure. For example, in binocular rivalry paradigms, the contrast is between the subject’s being in one specific conscious state (e.g. seeing a face) and another specific conscious state (e.g. seeing a house); similarly, in a masking paradigm, the contrast is between conscious and nonconscious content processing in participants who are conscious throughout the experiment. It seems plausible that some of the stable neuronal activity that is seen in these experiments—and which underwrites the fact that the participants are conscious—functions as a non-differentiating NCC for each of these specific con- scious experiences. Thus, there is reason to think that such studies are capable of revealing only the differentiating correlates of consciousness and are necessarily 2 Consciousness: Theoretical Approaches 27 unable to reveal the non-differentiating components of a conscious state’s full neural correlate (Hohwy 2009). What about the NCCs of the various modes of consciousness? Here there is an important distinction to be made between the modes of consciousness in which subjects retain robust access to their conscious states from those modes in which first-person access to consciousness is seriously impaired or even com- pletely lost. This latter category includes not just those modes that occur in disorders of consciousness (such as the minimally conscious state or epileptic absence seizures) but also certain modes that occur in healthy individuals, such as those seen in light anaesthesia or REM dreaming. In order to identify the NCC of these modes of consciousness, we need so-called ‘objective’ measures of consciousness, and there is notoriously little agreement about what such measures might look like (or indeed whether there even are any such measures that can be trusted). Finally, what about generic NCCs—the neural correlates of the state of being conscious as such? Because it is a trivial truth that any creature that is in a specific conscious state is itself conscious, any specific NCC will also be a generic NCC. However, there may also be generic NCCs that are not also specific NCCs. If there are such states, then we might be able to use them as markers of consciousness as such without needing to rely on specific NCCs. One method neuroscience can use to reveal generic NCCs is to form hypotheses about what all specific NCCs have in common (this would presumably include the non-differentiating NCCs) and then contrast conditions where this invariant part of NCCs is present with conditions where that part of the NCC is absent (i.e. states where unconsciousness is assumed, such as disorders of consciousness, some sei- zures, and anaesthesia). Without this contrast between wholly conscious and wholly unconscious creatures, it cannot be ruled out that a candidate for being the generic NCC is in fact unrelated to the presence of consciousness. Methodologically, this calls for manipulating both the contents of consciousness and the overall state of the creature, preferably in a full-factorial design. This is a non-trivial task because it is difficult to conceive of conditions that would yield either overall con- scious states without specific contents or overall unconscious states with specific contents (for discussion, see Bayne 2007; Hohwy 2009). This task is further com- plicated by the fact that the subject’s modes of consciousness may change during experimental manipulations and thereby affect the way in which their conscious contents are modulated. It is an open question whether there are such generic NCC states. Although there could be neural systems that are implicated in all forms of consciousness, it is also possible that consciousness is highly disjunctive from a neural point of view and that the neural correlates of different kinds of specific conscious states involve dis- tinct systems that may have little to nothing in common. ‘Global’ accounts of con- sciousness of the kind that have been defended by Baars and Dehaene (Baars 2005; Dehaene and Changeux 2011) would point towards the former possibility, whereas ‘local’ accounts of consciousness of the kind defended by Zeki (2007) and van Gaal and Lamme (2011) would point in the latter direction. 28 T. Bayne and J. Hohwy 2.4 The Ascription of Consciousness Arguably, an account of the NCCs would be of relatively little interest in its own right. Instead, its importance would reside in what we might be able to do with it. One idea that motivates much of the interest in the NCCs is the thought that we might be able to use an account of the NCCs as a tool for the ascription of consciousness. The markers for consciousness that we currently possess are problematic in various ways. Perhaps the most important of these markers are reports. We lean heavily on a person’s reports—or surrogates for such reports, such as button presses—in order to determine what conscious contents (‘I see a face’; ‘I smell coffee’) the person in question is enjoying. However, there are a number of respects in which verbal reports in general—and introspective reports in particular—have shortcomings when it comes to the ascription of consciousness. Firstly, many crea- tures are unable to produce any kind of report, let alone introspective reports. This is obviously true of most non-human animals and prelinguistic infants, but it also applies to many mature human beings who may have suffered from brain damage of some kind. Secondly, even creatures that are able to produce introspective reports may not always be reliable when it comes to describing or even detecting their own conscious states (Bayne and Spener 2010; Hohwy 2011; Schwitzgebel 2008). In some cases—as in emotional experiences—the content of the conscious state might be obscure and difficult to form an accurate judgement about. In other cases subjects form introspective judgements in conditions that are unlikely to be conducive to introspective reliability. A much-discussed example of this kind of problem is raised by Sperling displays, in which subjects are briefly presented with a grid of alphanumeric items. Although subjects are able to report only some of the figures, many subjects have the impression that they had been conscious of each of the presented items. As Ned Block (2007) has put it, many people have the impres- sion that the contents of consciousness ‘overflow’ what is accessible to the systems responsible for verbal report and short-term memory. Block and others have argued that in such cases subjects may have conscious states to which they lack introspec- tive access. So, we need ways of measuring the presence and absence of consciousness that don’t involve introspective report. In some contexts, behavioural measures of vari- ous kinds might provide us with tools for the ascription of consciousness. For exam- ple, clinicians employ a patient’s capacity for voluntary agency as a guide to the presence of consciousness when dealing with patients who have suffered serious brain trauma. However, even the most optimistic assessment of the power of non- verbal behaviour measures will grant that there are many contexts in which their capacity to provide us with an accurate account of a creature’s state of conscious- ness is questionable. It is tempting to hope that neuroscience in general and neuroimaging in particu- lar might provide us with some assistance here. If we could identify the NCCs, then—the thought is—we could use information about an individual’s neural states in order to identify its state of consciousness. It is unlikely that such measures would 2 Consciousness: Theoretical Approaches 29 supplant standard measures of consciousness that we currently employ, but they might be used when current methods are either silent or of dubious reliability. How might neuroimaging be used as a measure for the presence of conscious- ness? The most straightforward line of argument involves what is known as a ‘reverse inference’ (Poldrack 2006). Reverse inferences proceed in two steps. One first identifies the neural correlate (N1) for a certain kind of conscious state (C1) in a population. One then argues that if an arbitrary member of that population is in state N1, then there is good reason to ascribe C1 to that individual. The strength of this reason will depend on a number of things, most notably the robustness of the cor- relation between N1 and C1 in the relevant population. We will shortly examine a number of significant challenges to reverse inferences of this kind, but let us first note that at best arguments of this form will provide us with evidence that a creature is in a certain kind of conscious state. They cannot be used to show that a creature is not in a certain kind of conscious state. This is because the existence of robust mapping from N1 to C1 in a population is compatible with the lack of a tight mapping from C1 to N1 in that same population. For example, it could be that conscious state C1 is multiply realised in the relevant population and that in some individuals C1 is correlated with N1, whereas in other individuals C1 is correlated with neural states N2 and N3. If this were the case, then the fact that a member of this population was in N1 might give us good reason to think that it was also in C1, but the fact that it was not in N1 would not necessarily give us good rea- son to think that it was not in C1 (for the individual in question might be in either N2 or N3). So, inferences from the absence of an NCC to the absence of a conscious state can be precarious. (If we knew that the mapping from C1 to N1 was as robust as the mapping from N1 to C1, then the discovery that the creature was not in N1 would give us reason to think that it was not in N1, but this would involve knowing more than just that N1 is an NCC of C1.) Let us return to the challenges facing the use of reverse inferences to ascribe conscious states to an individual. We can explore the most pressing of these chal- lenges by considering the following striking study by Adrian Owen and colleagues, in which the brain of a vegetative-state patient was scanned while she was instructed to either imagine herself playing tennis or visiting the rooms of her home (Owen et al. 2006). They found that the fMRI signal in the areas that are preferentially implicated in these tasks (SMA for the tennis imagery and PMC/PPC/PPA for spa- tial navigation) was indistinguishable from that seen in 12 healthy controls. On the basis of this finding, the authors declared that there is little doubt that the patient was conscious, even though she was not at the time able to produce either introspec- tive reports or overt voluntary behaviour of any kind. This interpretation of the data can be—and indeed has been—challenged on a number of grounds. One question is whether the neural states appealed to in this study are full NCCs. Let us focus just on the relationship between SMA activity and conscious experiences of motor imagery. There is evidence that SMA activity is correlated with motor imagery in conscious individuals, but it doesn’t follow from this that SMA activity is itself sufficient for an experience of motor imagery. Instead, it could be the case that SMA activity is only a differentiating correlate of such 30 T. Bayne and J. Hohwy experiences and that a full correlate of such experiences may involve neural states of which SMA activity is only one component (the other component being the ‘non- differentiating’ correlate of consciousness). This is an important point here, because for all we know vegetative-state patients may not have the capacity for the neural activity required for non-differentiating correlates of consciousness. In short, SMA activity might constitute good evidence for experiences of motor imagery when dealing with conscious individuals, but it may not provide such evidence in the context of individuals in which the very presence of consciousness is uncertain. Even if SMA activity is a full correlate of conscious motor imagery in normal human beings, it is a further question whether SMA activity is correlated with con- scious motor imagery in this patient. The reason for this is that it is not clear whether this patient is a member of the population with respect to which this NCC holds. Is the relevant population for this NCC adult human beings, or is it adult human beings who have not suffered serious brain damage? The case for thinking that there is a correlation between SMA activation and experiences of motor imagery is based on studies of neurologically unimpaired individuals—individuals who are able to report their conscious states—but of course this patient is neither unimpaired nor is she able to report her conscious states. So, even if there is a robust correlation between SMA activation and experiences of motor imagery in neurologically unimpaired individu- als, it is an open question whether that correlation extends to individuals who have suffered significant brain damage. We might call this the ‘population problem’. The population problem might arise even when it comes to the ascription of con- sciousness to neurotypical individuals, for there is mounting evidence that individ- ual differences in perception and aspects of consciousness can be predicted by individual and local grey and white matter differences (Kanai and Rees 2011). If there are aspects of consciousness for which it is possible to identify only NCCs that are ‘individually tailored’ (or at least relativised to rather select populations), then the population problem will be even more acute, for it may be difficult to know which population to assign an arbitrarily selected individual to. A third challenge—one that is intimately related to that just discussed—concerns a worry about circularity. The problem is this: given that one must employ markers for the ascription of consciousness in order to discover the NCCs, how could the NCCs them- selves function as independent evidence of consciousness? One might think that the evidential power of an NCC could, of necessity, only be as strong as that of the markers that were used to identify it in the first place and thus that information about an indi- vidual’s neural state could never provide one with independent grounds for the ascrip- tion of consciousness. One might think that to the extent that there is a sound inference from an individual’s neural states to consciousness, there must also be a sound inference from some other property that that individual possesses (e.g. a behavioural property) to consciousness. In other words, as far as the ascription of consciousness goes, the worry is that neuroimaging data is either ungrounded or redundant. This last challenge raises what is perhaps the most acute challenge for any attempt to use neuroimaging to ascribe consciousness, and the jury is still out on whether it can be given a satisfactory response. The most promising responses to it invoke inference to the best explanation, a pattern of justification that is ubiquitous 2 Consciousness: Theoretical Approaches 31 in science. The idea is that we might be justified in regarding a certain type of neural state N1 as a more robust marker of any set of pre-theoretical markers of consciousness—markers that were employed in identifying N1—on the grounds that N1 provides a unifying account of why those markers tend to co-occur in the way that they do (Shea 2012). Thus far we have focused on direct forms of the reverse inference argument, but the reverse inference model can also be deployed in less direct ways. One might argue that neuroimaging data provides evidence of consciousness in virtue of provid- ing evidence of some cognitive capacity which is itself good evidence of conscious- ness. For example, one might argue that the fMRI data obtained by Owen and colleagues is evidence of consciousness insofar as it is evidence of intentional agency (it was sustained for 30 s and was time-locked to the imagery instructions given to the patient) and intentional agency is itself evidence for consciousness (Shea and Bayne 2010). According to this reconstruction of the argument, neuroimaging data justifies an inference to consciousness not because it concerns particular neural areas but because of its duration and relationship to the patient’s environment. This indirect version of the argument is also open to challenge. For one thing, some theorists have argued that the activation seen in this patient was not a manifes- tation of intentional agency but was merely automatically triggered, even though it lasted for 30 s and was time-locked to the experimental instructions (Naccache 2006). Even if it is granted that the patient was engaged in an intentional action, one might argue that it is a further question whether intentional agency is a reliable indicator of consciousness (Levy 2008). A form of the population problem arises here too, for even if the presence of intentional agency is correlated with conscious- ness in the healthy population, such a correlation might not hold when it comes to individuals with disorders of consciousness (Hohwy and Fox 2012). It goes beyond the scope of this chapter to evaluate these important but difficult issues further. Our aim is only to draw attention to the fact that there is more than one way in which one might mount a case for consciousness on the basis of neuroimaging data. 2.5 The Explanation of Consciousness We turn now to the question of whether an account of the NCCs might point us in the direction of an explanation of consciousness. It is one thing to know that a cer- tain kind of neural state is correlated with a certain kind of conscious state, but can we go beyond correlations to explanations? There is very good reason to think that certain kinds of mental phenomena can be explained in neural terms. In memory research, for example, neuroscience has uncovered neural mechanisms such as long-term potentiation that explain how functional features of memory arise from brain activity. An understanding of these mechanisms provides us with a reductive account of memory—it removes any of the mystery surrounding memory (Craver 2007). Might an account of the NCCs also provide—or at least point the way towards providing—a reductive explanation of consciousness? Might it remove the mystery surrounding consciousness? 32 T. Bayne and J. Hohwy There are reasons for skepticism. To use Joseph Levine’s (1983) famous phrase, there seems to be an explanatory gap between neural states and conscious states. Information about the neural states that underlie consciousness seems unable to explain why such states are accompanied by conscious states. Assume that we have identified not only neural correlates for every aspect of consciousness but also neu- ral mechanisms that explain every functional and structural feature of conscious- ness. Such an explanation would seem to leave something unexplained—it would seem to leave an explanatory gap. In fact, there are two kinds of explanatory gaps to grapple with here: a generic gap and a specific gap. The generic gap concerns the question of why some neural states are accompanied by conscious states of any kind, while the specific gap concerns the question of why N1 is accompanied by the specific kind of conscious state that it is (say, the smell of coffee) rather than another (say, the visual experience of a face). Why is there ‘something it’s like’ to be in N1, and why is what it’s like to be in N1 like this rather than like that? Neural states appear able to explain only the structural and functional features of mental phenom- ena, and yet an account that appeals only to structure and function seems doomed to omit the ‘phenomenal feel’ of consciousness. On the face of things, a robot could exemplify the functional and structural aspects of consciousness without there being anything ‘that it’s like’ for the robot to be the robot it is. Broadly speaking, there are four kinds of responses to the explanatory gap in the literature. Some theorists (e.g. Dennett 1991) deny that there is anything more to consciousness than structure and function and thus hold that neuroscience possesses the conceptual tools required to explain every aspect of consciousness. Other theo- rists (e.g. Searle 2004) allow that consciousness involves more than structure and function, but they hold that neuroscience will develop the conceptual tools needed in order to explain these additional aspects of consciousness. A third group of theo- rists (e.g. McGinn 1989) hold that there is a perfectly natural account of how con- sciousness emerges from neural activity, but they hold that our cognitive limitations prevent us from grasping that account. A final group of theorists (e.g. Chalmers 1996) argue that the explanatory gap is a manifestation of an underlying metaphysi- cal gap between consciousness and the physical world: neuroscience will not be able to explain how consciousness emerges from neural activity because conscious- ness doesn’t emerge from neural activity—it is merely correlated with it. The debate between these four positions cuts to the heart of some of the deepest and most obscure questions in philosophy, and we are not minded to engage with it here. Even if a reductive treatment of consciousness in its entirety is beyond our ken, neuroscience may be able to provide reductive explanations of those aspects of consciousness that do involve structural and functional properties. For example, by providing an account of how neural states might implement certain computational states, we can begin to see how it might be possible to explain the relationship between consciousness and other mental phenomena such as attention, memory, intention, and reasoning. We will bring this chapter to a close by considering certain aspects of this explanatory project. One aspect of consciousness that may well succumb to reductive explanation is its modal nature. As we noted earlier, the various modes of consciousness differ 2 Consciousness: Theoretical Approaches 33 from each other in terms of the cognitive and behavioural capacities with which they are associated. For example, various disorders of consciousness (such as those seen in epileptic absence seizures) are associated with reduced availability of the contents of consciousness to the ‘high-level’ consuming systems involved in rea- soning, intentional control, and introspection. There is every reason to expect that a detailed analysis of the neural correlates of consciousness will point the way towards mechanistic accounts of why the various modes of consciousness are associated with their distinctive cognitive and behavioural profiles. Such an account might not bridge the explanatory gap, but it would mark a very considerable advance in our understanding of the nature of consciousness. By revealing the neural mechanisms underlying this aspect of consciousness, we may learn something new about con- sciousness—something that introspection would not be able to reveal. Recent research in this area is beginning to make some progress. For example, using Bayesian model selection, Boly and colleagues provide evidence that top- down message passing in the brain is implicated in the contrast between the modes of consciousness seen in disorders of consciousness and those seen in the state of normal wakefulness (Boly et al. 2011). Research like this provides a promising glimpse of the kind of mechanism that might constitute such modes of conscious- ness. Moreover, large-scale, top-down modulation of neural activity seems to be the kind of mechanistic process that could begin to explain the generation of an overall conscious state because its functional profile fits with the idea that consciousness involves integrating a number of low-level processing streams under a single, unified global model of the world. What we begin to see here is a marriage of neuroimaging findings from the search for the NCC with computational and information theoretical approaches to the neural systems making up NCCs (e.g. in terms of information integration (Tononi 2005), which is being incorporated into the global neuronal workspace theory (Dehaene and Changeux 2011) or prediction error minimisation (Friston and Stephan 2007)). This means that the neural activity correlating with conscious states can begin to be under- stood in its own right, without being picked out only via its correlation with con- scious states. This in turn improves the chances of identifying systematic NCCs where we can use our new theoretical, mechanistic understanding of the NCC to predict its behaviour under new types of experimental interventions on both contents, modes, and overall state. For example, using transcranial magnetic stimulation of NCCs, based on our understanding of its computational properties, it may be possi- ble to produce new types of conscious states. This is something that cannot be done as long as the NCC is exclusively picked out as the correlates of conscious states with no understanding of their mechanistic and computational function. In his influential treatment of the NCCs, Chalmers (2000) noted the desirability of identifying systematic correlations between neural states and consciousness rather than mere ‘raw’ or ‘one-off’ correlates (2000). Contemporary computational and information theoretical work is paving the way for the realisation of this aim. There is reason to believe that the combination of traditional NCC approaches for contents, modes, and generic conscious states with computational theory will be a main contributor to our understanding of how these aspects of consciousness interact 34 T. Bayne and J. Hohwy with each other. There is thus considerable reason to think that neuroscience will be able to explain many of the structural and functional aspects of consciousness. We may not be able to close the explanatory gap, but we will certainly be able to narrow it. References Alkire MT, Hudetz AG et al (2008) Consciousness and anesthesia. Science 322(5903):876–880 Baars BJ (2005) Global workspace theory of consciousness: toward a cognitive neuroscience of human experience. Prog Brain Res 150:45–53 Bayne T (2007) Conscious states and conscious creatures: explanation in the scientific study of consciousness. Philos Perspect 21:1–22 Bayne T, Montague M (2011) Cognitive phenomenology. Oxford University Press, Oxford Bayne T, Spener M (2010) Introspective humility. Philos Issues 20:1–22 Bekinschtein TA, Dehaene S et al (2009) Neural signature of the conscious processing of auditory regularities. P Natl Acad Sci (USA) 106(5):1672–1677 Block N (2007) The methodological puzzle of consciousness and the mesh between psychology and neuroscience. Behav Brain Sci 30:481–548 Boly M, Garrido MI et al (2011) Preserved feedforward but impaired top-down processes in the vegetative state. Science 332(6031):858–862 Chalmers D (1996) The conscious mind. Oxford University Press, New York Chalmers D (2000) What is a neural correlate of consciousness? In: Metzinger T (ed) The neural correlates of consciousness. MIT Press, Cambridge, pp 17–39 Craver C (2007) Explaining the brain: mechanisms and the mosaic unity of neuroscience. Oxford University Press, New York Dehaene S, Changeux J-P (2011) Experimental and theoretical approaches to conscious processing. Neuron 70(2):200–227 Dennett DC (1991) Consciousness explained. Little, Brown & Co., Boston Friston K, Stephan K (2007) Free energy and the brain. Synthese 159(3):417–458 Hohwy J (2007) The search for neural correlates of consciousness. Philos Compass 2(3): 461–474 Hohwy J (2009) The neural correlates of consciousness: new experimental approaches needed? Conscious Cogn 18:428–438 Hohwy J (2011) Phenomenal variability and introspective reliability. Mind Lang 26:261–286 Hohwy J, Fox E (2012) Preserved aspects of consciousness in disorders of consciousness: a review and conceptual analysis. J Consciousness Stud 19(3–4):87–120 Kanai R, Rees G (2011) The structural basis of inter-individual differences in human behaviour and cognition. Nat Rev Neurosci 12(4):231–242 Laureys S (2005) The neural correlate of (un)awareness: lessons from the vegetative state. Trends Cogn Sci 9(12):556–559 Levine J (1983) Materialism and qualia: the explanatory gap. Pac Philos Quart 64:354–361 Levy N (2008) Going beyond the evidence. Am J Bioeth 8(9):19–21 McGinn C (1989) Can we solve the mind–body problem? Mind 391:349–366 Naccache L (2006) Is she conscious? Science 313:1395–1396 Nagel T (1974) What is it like to be a bat? Philos Rev 83:435–450 Owen AM, Coleman MR, Boly M et al (2006) Detecting awareness in the vegetative state. Science 313:1402 Poldrack RA (2006) Can cognitive processes be inferred from neuroimaging data? Trends Cogn Sci 10(2):59–63 Schwitzgebel E (2008) The unreliability of naïve introspection. Philos Rev 117:245–273 Searle JR (2004) Mind – a brief introduction. Oxford University Press, Oxford 2 Consciousness: Theoretical Approaches 35 Shea N (2012) Methodological encounters with the phenomenal kind. Philos Phenomen Res 84:307–344 Shea N, Bayne T (2010) The vegetative state and the science of consciousness. Brit J Philos Sci 61:459–484 Tononi G (2005) Consciousness, information integration, and the brain. Prog Brain Res 150:109 van Gaal S, Lamme VAF (2011) Unconscious high-level information processing: implication for neurobiological theories of consciousness. Neuroscientist. doi:10.1177/1073858411404079 Zeki, S (2007) A theory of micro–consciousness. In M Velmans and S Schneider (Eds) The Blackwell companion to consciousness. Blackwell