1. Introduction

There is a consensus that conscious states are associated with only a subset of the many sophisticated processes that have been identified in the human nervous system (Baars, 2002; Crick & Koch, 2003; Dehaene & Naccache, 2001; Gray, 2004; Merker, 2007; Morsella, Krieger, & Bargh, 2010). By 'conscious state' we are referring to the most basic form of consciousness, the kind of consciousness that has fallen under the rubrics of 'basic awareness,' 'sentience,' and 'phenomenal state.' This most basic form of consciousness has been defined best by the philosopher Nagel (1974), who claimed that an organism has phenomenal states if there is something it is like to be that organism--something it is like, for example, to be human and experience pain, love, breathlessness, or yellow afterimages.

The conscious state is 'everything' to us, because it encompasses the totality of our human experience. However, knowledge of nervous function reveals that, normally unbeknownst to us, many of the complicated functions in the nervous system are carried out beneath the horizon of basic consciousness. For example, unconscious processes include (a) low-level perceptual processing, such as the putting together of perceptual features both within and across sensory modalities (e.g., vision and touch), and (b) motor control, as in the control of the impulses that contract some muscle fibers but not others when carrying out an action. Moreover, sophisticated processes such as those constituting that syntax and the parsing of sentences are largely unconscious. Appreciating all that can be achieved unconsciously in the brain leads one to the question, What do conscious states contribute to nervous function? To begin to answer this question, it is helpful as a first step to isolate the cognitive/brain processes that seem to be most intimately associated with conscious states.

We will first present the results of investigations seeking to isolate the cognitive processes that are most associated with consciousness; then we will review data isolating consciousness to a subset of brain processes.

2. The Subset of Basic Cognitive Processes Associated with Conscious States

Most cognitive operations occur unconsciously. For example, there is substantial evidence for the unconscious nature of low-level perceptual analysis and of the integration of sensory information within a sensory modality, such as the 'binding' of shape and color of an orange (Zeki & Bartels, 1999), and across modalities, as in the countless audiovisual integrations responsible for illusions such as the ventriloquism effect and the classic and dramatic McGurk effect (McGurk & MacDonald, 1976). (In this effect, an observer views a speaker mouthing 'ga' while presented with the sound 'ba.' Surprisingly, the observer is unaware of any intersensory interaction, perceiving only 'da.') Regarding semantics (e.g., the meaning of words), evidence suggests that many of its workings, too, are unconscious. For example, a speaker does not know that, when naming a cat, the meaning of the concept DOG was activated, at least to some extent (Levelt, 1989).

Regarding action, there is evidence that action plans can be activated, selected, and even expressed unconsciously (e.g., during neurological disorders; see review in Morsella & Bargh, 2011). This is obvious in actions such as the pupillary reflex, reflexive pain withdrawal, and in behavioral responses to stimuli that have been rendered subliminal through techniques such as 'visual masking' (Hallet, 2007). Convergent evidence for the existence of unconscious action is found in neurological cases where, following brain injury in which a general awareness is spared, actions can are decoupled from consciousness, as in blindsight (Weiskrantz, 1997), in which patients report to be blind but still exhibit visually guided behaviors. Similarly, in alien hand syndrome (Bryon & Jedynak, 1972), anarchic hand syndrome (Marchetti & Della Sala, 1998), and utilization behavior syndrome (Lhermitte, 1983), brain damage causes hands and arms to function autonomously. These actions include relatively complex goal-directed behavior (e.g., the manipulation of objects; Yamadori, 1997) that are maladaptive and, in some cases, can be at odds with a patient's reported intentions (Marchetti & Della Sala, 1998). In addition, Goodale and Milner (2004) report neurological cases in which there is a dissociation between action and conscious perception. Patient D.F., suffering from visual form agnosia, was incapable of reporting the orientation of a tilted slot, but could nonetheless negotiate the slot accurately when inserting an object into it. Complex integrations that are wholly unconscious occur regularly in motor control (Grossberg, 1999; Jeannerod, 2006; Rosenbaum, 2002) and in the control of smooth muscle (e.g., the pupillary reflex; Bartley, 1942; Morsella, Gray, Krieger, & Bargh, 2009).

Which processes tend to always be associated with conscious states? According to the integration consensus (Morsella, 2005), conscious states appear to furnish the nervous system with a form of internal communication that integrates neural activities and information-processing structures that would otherwise be independent, permitting diverse kinds of information to be gathered in some sort of global workspace, thus allowing adaptive action to emerge (cf., Baars, 2002; Dehaene & Naccache, 2001; Merker, 2007; Morsella, 2005). The kinds of unconsciously-mediated actions described above lack this form of integration and thus appear irrational and impulsive, meaning that the actions are not constrained by information that, for adaptive action, should be influencing them. But exactly which kind of integration requires the conscious state? As mentioned above, many integrational processes (e.g., intersensory illusions) occur in the nervous system unconsciously.

To address this issue, Supramodular Interaction Theory (SIT; Morsella, 2005) proposes that conscious states are necessary to integrate diverse sources of information/processes, but only certain kinds of information/processes. For example, conscious states are unnecessary to integrations between perceptual processes, as in afference binding (e.g., intra- or inter-sensory interactions; Morsella & Bargh, 2011). Similarly, basic stimulus-response (S -> R) associations (efference binding; Haggard, Aschersleben, Gehrke, & Prinz, 2002), such as inhaling reflexively or pressing a button in response to a subliminal stimulus (Fehrer & Biederman, 1962; Fehrer & Raab, 1962; Taylor & McCloskey, 1990, 1996) can occur unconsciously. (See review of correct motor responses to subliminal stimuli in Hallett, 2007.) Figure 1 reviews all the many unconscious forms of interaction (or 'crosstalk') in the brain.

According to SIT, the difference between conscious and unconscious interactions is not simply about the complexity, controllability, or feedback properties of the processes involved, nor is it about the degree to which the processes require memory, semantics, action-related mechanisms, or 'top-down' processes (see extensive treatment of this in Morsella, 2005). Rather, what most reliably distinguishes conscious from unconscious interactions pertains to the nature of the effectors involved (Morsella, 2005).

For example, research reveals that conscious conflicts (e.g., holding one's breath or suppressing a pain-withdrawal reflex) are special in that they involve conflicting tendencies toward the skeletal muscle output system. Specifically, consciousness is reliably perturbed when there is the simultaneous activation of two conflicting streams of efference binding toward the skeletal muscle output system (e.g., signaling inhale and do not inhale when holding one's breath underwater; see evidence in Morsella, Berger, & Krieger, in press). Such efference-efference binding results in integrated actions (Morsella & Bargh, 2011) such as holding one's breath, breathing faster for a reward, carrying a hot dish of food, or suppressing socially-inappropriate behavior (Figure 2).

Conscious states are not necessary for intersensory conflicts, smooth muscle actions, or skeletomotor actions that are unintegrated (Morsella & Bargh, 2011), that is, those driven by a single stimulus-response (S -> R) stream, such as withdrawing one's hand from a hot stove or responding to a subliminal stimulus. SIT is unique in its ability to explain the subjective effects of conflicts from action conflicts (e.g., holding one's breath) and the lack of subjective effect from intersensory conflicts or smooth muscle conflicts. From this standpoint, the skeletal muscle system is a multi-determined effector controlled by many brain systems, each potentially having distinct phylogenetic origins and operating principles. This form of multi-determination reveals that action selection at the organismic level suffers from a 'degrees of freedom' problem (Rosenbaum, 2002): There are simply too many things that one can decide to do next. For instance, during the trial of a laboratory experiment about action conflict, a subcortical brain region may want the eyes to glance leftwards (because a bright flash occurred there), and a cortical region may want the eyes to look rightwards (because of the experimenter's instruction). The challenge of multidetermination in action selection is met, not by unconscious motor algorithms (as in the case of motor control; Rosenbaum, 2002), but by the ability of conscious states to constrain what the organism does by having the inclinations of multiple systems co-exist in the conscious field and thereby constrain skeletomotor output.

Figuratively speaking, the skeletal muscle system is like a big steering wheel that different parts of the brain are trying to control at the same time (Morsella, Krieger, & Bargh, 2009). When these different systems are in conflict, as when one voluntarily holds one breath--and part of us wants to inhale and another part of us does not want to inhale--we have a strong conscious experience. Consciousness is the way the different regions communicate with each other. Without consciousness, there would be no crosstalk between the different systems and one would inhale reflexively. One does not need consciousness to withdraw from a painful stimulus, but one does need it in order to keep touching it. Another way to think about it is as follows. There are many quasiindependent computers in the brain, and each can do complicated things and influence overt action, which we can think of as influencing the actions of a printer. Each computer can influence the printer, but in order for two computers to interact and then influence the printer, you need a wifi system. Consistent with the integration consensus (stating that consciousness integrates information/processes that would be independent otherwise), in this way consciousness functions as a wifi system to integrate the different processes in the brain and yield adaptive action. One can certainly imagine integration among systems occurring without anything like a conscious state, but in this descriptive approach, and for the reasons that only the tinkering and happenstance process of evolution could explain, it was these states that were selected to solve this particular integration problem in the brain (Morsella, 2005). SIT is unique in also explaining why skeletal muscle is voluntary muscle: Skeletomotor actions are at times 'consciously mediated' because they are directed by multiple, encapsulated systems that, when in conflict, require consciousness to yield adaptive action.

Through the process of elimination (i.e., identifying all the cognitive operations that the nervous system can carry out unconsciously), we have attempted to identify what conscious states are for by isolating the few basic processes that seem to be intimately related with these states and that seem incapable of occurring unconsciously. Can a similar eliminative approach isolate the neural substrates of conscious states?

3. The Subset of Brain Areas/Processes Associated with Conscious States

When homing in on the conscious state neuroanatomically, the evidence not as straightforward as the evidence for isolating the function of conscious states in processing, partly because consciousness may not so much be associated with a specific brain region as with a particular mode of processing among regions. It seems that the mode of interaction among regions is as important as the nature and loci of the regions (Buzsáki, 2006). For example, the presence or lack of interregional synchrony leads to different cognitive and behavioral outcomes (Hummel & Gerloff, 2005; see review of neuronal communication through 'coherence' in Fries, 2005). That the mode of interaction between areas is important for conscious states is evident in binocular rivalry. In binocular rivalry (Logothetis & Schall, 1989), an observer is presented with different visual stimuli to each eye (e.g., an image of a house in one eye and of a face in the other).

It might seem reasonable that, faced with such stimuli, one would perceive an image combining both objects--a house overlapping a face. Surprisingly, however, an observer experiences seeing only one object at time (a house and then a face), even though both images are always present. At any moment, the observer is unaware of the computational processes leading to this outcome; the conflict and its resolution are unconscious. Consistent with the view that the mode of interaction between areas is important for consciousness, during binocular rivalry, it is only the conscious percept that, neurally, is coupled to both perceptual brain activity and motor-related processes in frontal cortex (Doesburg, Green, McDonald, & Ward, 2009)

Unconscious processes involve smaller networks of brain areas than their conscious counterparts (Gaillard et al., 2009; Sergent & Dehaene, 2004; See review in Morsella, Krieger, & Bargh, 2010.) For example, unconsciously-mediated actions (e.g., reflexive pharyngeal swallowing) involve substantially fewer brain regions than their voluntary counterparts (e.g., volitional swallowing; Kern, Jaradeh, Arndorfer, & Shaker, 2001; Ortinski & Meador, 2004). Regarding gross neuroanatomy, consciousness has been linked to the 'ventral processing stream' of the brain, which is not necessary for action execution but for knowledge-based action selection (Goodale & Milner, 2004). It has been proposed that conscious states require a form of thalamocortical interaction (or resonance) between thalamic 'relay' neurons and cortical neurons (Coenen, 1998; Edelman, Baars, & Seth 2005; Llinás & Ribrary, 2001; Ojemann, 1986), but this is inconsistent with the fact that we consciously experience aspects of olfaction even though the afferents from the olfactory sensory system bypass the thalamus and directly target regions of the ipsilateral cortex (Morsella, Krieger, & Bargh, 2010; Shepherd & Greer, 1998). This is not to imply that conscious olfaction does not require the thalamus: in post-cortical stages of processing, the thalamus receives inputs from cortical regions that are involved in olfactory processing (Haberly, 1998). Buck (2000) proposes that conscious aspects of odor discrimination depend primarily on the activities of the frontal and orbitofrontal cortices; Barr and Kiernan (1993) propose that olfactory consciousness depends on the pyriform cortex. These proposals appear inconsistent with subcortical accounts of consciousness (Merker, 2007; Penfield & Jasper, 1954).

Regarding the cerebral cortex, experiments on 'split-brain' patients (Wolford, Miller, & Gazzaniga, 2004), binocular rivalry (Logothetis & Schall, 1989), and splitbrain patients experiencing binocular rivalry (O'Shea & Corballis, 2005) strongly suggest that basic consciousness does not require the non-dominant (usually right) cerebral cortex nor the commissures linking the two cortices. Evidence also suggests that, although the absence of the spinal cord or cerebellum leads to sensory, motor, cognitive, and affective deficits, it does not seem to eradicate the basic conscious state (Schmahmann, 1998).

Similarly, although extirpation of the amygdalae or hippocampi leads to anomalies including severe deficits in affective processing (LeDoux, 1996) and episodic memory (Milner, 1966), respectively, it seems that a basic conscious state persists without these structures. Investigations regarding prefrontal lobe syndromes (Gray, 2004), the phenomenology of action (Desmurget et al., 2009; Desmurget & Sirigu, 2010), and the psychophysiology of dream consciousness, which involves prefrontal deactivations (Muzur, Pace-Schott, & Hobson, 2002), suggest that, although the prefrontal lobes are involved in cognitive control (see review in Miller, 2007), they are not essential for the generation of basic consciousness. According to Gray (2004), one is conscious, not of high-level executive processes or motor efference, but only of perceptual-like contents, whether they precede an action, as in the case of anticipated action effects, or whether they follow an action, as in the case of experiencing the actual action effects. This is consistent with recent experiments by Sirigu and colleagues. Research from the Sirigu laboratory (Desmurget et al., 2009; Desmurget & Sirigu, 2010) reveals that direct electrical stimulation of parietal areas of the brain gives subjective urges and that increased activation makes them believe that they actually executed the corresponding action, even though no action was performed. Motor activations (e.g., in premotor areas) can lead to the actual action, but subjects believe that they did not perform any action. Interestingly, this approach is consistent with what ideomotor approaches (Hommel, 2009; Hommel, Müsseler, Aschersleben, & Prinz 2001) to action control have proposed: That awareness of our actions occurs only in a 'sensorium' (a term used by Johannes Müller) with motor processes being unconscious.

In contrast to cortical accounts of consciousness, Penfield and Jasper (1954) proposed that these states are primarily a function of subcortical structures. This 'cortical-subcortical' controversy arose from Penfield and Jasper's (1954) observations of awake patients undergoing brain surgeries involving ablations and direct brain stimulation. Penfield and Jasper concluded that, though the cortex may elaborate the contents of consciousness, it is not the seat of consciousness. Recently, based on clinical observations of anencephaly, Merker (2007) re-introduces this hypothesis in a theoretical framework in which consciousness is primarily a phenomenon associated with mesencephalic areas. It seems reasonable to conclude that consciousness can persist even when great quantities of the cortex are absent (Merker, 2007). The outstanding question is whether an identifiable form of consciousness can exist despite the non-participation of all cortical matter. Cortical areas that seem most associated with the conscious state are the sensory and parietal areas. Future research on the cortical-subcortical controversy may evaluate the necessary role of parietal areas (e.g., those identified by Desmurget and Sirigu, 2010) in the generation of basic consciousness.

4. The Place of Consciousness in the Physical World

In summary, isolating the cognitive underpinnings of conscious states (some kind of integrative process for adaptive skeletomotor function) has proven easier than isolating the neural underpinnings.

We adopted an inductive and descriptive approach, in which nervous function is described as is, and not as it (perhaps) should be--what would be a normative rather than a descriptive approach. Hence, intuitions regarding how the nervous system should work take a back seat to actual data revealing how it actually works, whether optimally or suboptimally. Thus, it is premature to propose that these states are 'epiphenomenal,' that is, serving no function whatsoever. Until one understands the place of a given phenomenon in nature, and how it emerges from nature, one should not make the strong claim that the phenomenon be epiphenomenal. Second, the current approach may identify what these states are for, but it sheds no light on why 'subjectivity' is associated with the tightly circumscribed, integrative function that these states appear subserve.