How does this theory account for known dissociations between consciousness and neural processing — such as anesthesia abolishing consciousness while preserving neural activity, split-brain cases, and blindsight?

IIT provides a principled explanation for the loss of consciousness during anesthesia and deep sleep by appealing to the breakdown of integrated information. During slow-wave sleep, changes in neuromodulation cause neurons to become bistable, rapidly alternating between activation and a stereotypical OFF-period lasting hundreds of milliseconds. This bistability produces a collapse of intrinsic information --- since most inputs lead to the same stereotypical output (the OFF-period) --- and consequently a collapse of integrated information. Experiments with epilepsy patients have confirmed that while electrical stimulation of the cortex during wakefulness triggers a chain of phase-locked activations, during slow-wave sleep the same input triggers a stereotyped slow wave associated with a cortical OFF-period, after which cortical activity resumes but the phase-locking is lost. This empirical finding has been generalized through the perturbational complexity index (PCI), a TMS-EEG-based measure that serves as a proxy for integrated information. PCI has been shown to be invariably high when subjects are conscious --- whether awake or dreaming --- and low when consciousness is lost in dreamless sleep, under various anesthetic agents (including ketamine, propofol, and midazolam), and during generalized seizures. The key insight is that anesthesia and deep sleep abolish consciousness not by eliminating neural activity per se, but by disrupting the effective connectivity among cortical neurons, reducing the brain's capacity to support a large, irreducible cause-effect structure.

Regarding split-brain cases, IIT's exclusion postulate predicts that the main complex must correspond to a definite set of units with maximal integrated information. After surgical resection of the corpus callosum, the theory accounts for the splitting of consciousness by attributing the phenomenal split primarily to the severing of posterior-central callosal connections. An intriguing prediction that follows from the exclusion postulate is that if one could progressively reduce the efficacy of callosal transmission, there would be a critical moment at which a single consciousness would suddenly split into two, following a minor change in the traffic of neural impulses across the callosum. This is because the exclusion postulate demands that the main complex be maximally irreducible --- a gradual reduction in callosal connectivity could reach a threshold where two separate cortical hemispheres each constitute a higher local maximum of Phi than the combined system, triggering an abrupt transition from one consciousness to two.

IIT also offers a framework for understanding blindsight and related dissociative phenomena. In blindsight, patients are subjectively blind in part of their visual field but can still make accurate forced-choice judgments about stimuli presented there. IIT can account for this through several mechanisms. First, the theory predicts that early sensory areas whose neurons specify causes and effects that are incongruent with the cause-effect state of the main complex as a whole may not contribute to experience. If low-level visual areas are damaged but some high-level areas remain intact and capable of being activated through subcortical pathways, patients may experience abstract invariants (such as a vague feeling of movement) without the low-level features normally composing a full visual experience. More broadly, IIT predicts that a dissociation from the main complex can occur in functional blindness: a patient may be subjectively blind but capable of purposefully avoiding obstacles, because the relevant neural processes may operate outside the main complex, as part of minor complexes or through residual interactions that do not contribute to the dominant conscious experience.

Global Workspace Theory provides detailed and largely convergent accounts of all three dissociations. Regarding general anesthesia, Baars (2005) notes that four causally very different unconscious states — deep sleep, coma, vegetative states, epileptic loss of consciousness, and general anesthesia under various agents — share a common set of features predicted by the theory: widely synchronized slow waveforms that replace the fast and flexible interactions needed for conscious functions, frontoparietal hypometabolism, widely blocked functional connectivity (both corticocortical and thalamocortical), and behavioral unresponsiveness to normally conscious stimuli. PET imaging shows marked hypofunction in frontoparietal association areas across all four conditions compared to waking consciousness. EEG studies reveal that at loss of consciousness under anesthesia, gamma power decreased while lower frequency bands increased especially in frontal leads, accompanied by a significant drop in coherence between homologous areas of the two hemispheres and between posterior and anterior regions. From the workspace perspective, this is interpreted as a loss of the long-range coordination between frontal and posterior cortex that is necessary for global broadcast — the anesthetic disrupts the workspace architecture not by silencing all neural activity but by severing the functional connectivity that allows local processing to be amplified and shared brain-wide.

For split-brain cases, Baars (2005) discusses Gazzaniga's findings that callosotomy patients encounter conflict between right and left hemisphere executive and perceptual functions. The left hemisphere's "narrative interpreter" in prefrontal cortex receives its own flow of sensory input and attempts to maintain coherent expectations and intentions, sometimes confabulating to repair its understanding of behavior driven by the disconnected right hemisphere. Baars suggests that full consciousness "may not exist without the participation of such prefrontal self systems," and that the left-hemisphere narrative interpreter can be viewed as a higher-level context system within GW theory. Dehaene and Naccache (2001) further note that in split-brain subjects, the left-hemispheric verbal "interpreter" invents plausible but false explanations for behavior caused by the right hemisphere, illustrating how workspace access is tied to the ability to report. The theory predicts that severing the corpus callosum disrupts the global workspace by fragmenting the long-distance connectivity required for unified broadcast, potentially creating two partially independent workspaces within each hemisphere.

Blindsight receives the most thorough treatment in both papers. Dehaene and Naccache (2001) observe that blindsight patients with lesions to visual cortical areas can detect and even point to visual stimuli in their blind field while denying conscious perception. Crucially, the theory notes that this does not make them "zombies": they never spontaneously initiate visually-guided behavior in the affected field and good performance can only be elicited by forcing them to respond. The workspace account holds that the lesion destroys or weakens the cortical processors that would normally be amplified into the workspace, so the information may still propagate through subcortical circuits (such as the superior colliculus) or residual cortical pathways — sufficient for above-chance forced-choice performance but insufficient to establish a sustained closed loop with workspace neurons. The information remains permanently or transiently inaccessible to the global workspace and therefore to consciousness, while local modular processing continues to support implicit behavioral responses.

HOT theory provides a natural framework for understanding blindsight, which Rosenthal's account addresses most directly among these dissociations. In blindsight, patients with damage to primary visual cortex retain the ability to make accurate forced-choice discriminations about visual stimuli in their scotoma, yet report no conscious visual experience. On the HOT account, this is precisely what one would expect when first-order visual processing continues but the pathway generating higher-order thoughts about those visual states is disrupted. The patient has genuine first-order visual states -- states with intentional content about features of the visual world -- but these states are not accompanied by higher-order thoughts representing the patient as currently having visual experiences. The visual states influence behavior (enabling above-chance performance) just as any nonconscious mental state might, but in the absence of higher-order representation they lack the subjective quality of conscious seeing. Rosenthal's broader argument that many mental states routinely occur outside the stream of consciousness -- beliefs, desires, expectations, and even bodily sensations like headaches that temporarily leave awareness when attention shifts -- provides the theoretical basis for this account. Blindsight is simply a pathological instance of the general phenomenon of nonconscious mental processing.

Regarding anesthesia, while Rosenthal does not directly discuss anesthetic agents in the 1986 paper, the theory's structure generates clear predictions. If consciousness requires that first-order states be accompanied by higher-order thoughts, then anesthesia could abolish consciousness by disrupting the generation of higher-order thoughts while leaving first-order sensory processing relatively intact. The theory predicts that anesthetic agents should particularly affect the neural systems responsible for producing higher-order representations -- plausibly higher-order cortical areas involved in metacognitive processing -- rather than necessarily eliminating all neural activity. Sensory neurons might continue to respond to stimulation under anesthesia, but if the cognitive machinery for generating higher-order thoughts is suppressed, no conscious experience would result. This fits the known empirical profile of anesthesia reasonably well, though the prediction remains generic since the theory does not specify which neural populations implement the higher-order thought mechanism.

For split-brain cases, the theory suggests that the unity of consciousness depends on the integrated operation of the higher-order thought system. In Rosenthal's framework, a higher-order thought must represent that one is, oneself, in a particular mental state, and this self-referential structure supports a sense of unified conscious experience. When commissural connections are severed, each hemisphere could retain the capacity to generate higher-order thoughts about its own first-order states while losing access to states processed exclusively by the other hemisphere. This would predict precisely the kind of dissociation observed in split-brain patients: each hemisphere maintains its own sphere of conscious experience. Rosenthal's discussion of how a sense of unified selfhood emerges from the accumulation of self-referential higher-order thoughts over time further supports this interpretation -- disrupting the integration of these thoughts should disrupt the unity of consciousness. However, these applications to anesthesia and split-brain remain extrapolations from the theory's principles rather than detailed accounts worked out in the source paper itself.

Regarding anesthesia and disorders of consciousness, PP offers a principled account based on the disruption of hierarchical predictive processing. Hohwy (2012) cites research showing that what is required for an individual to be in an overall conscious state is intact connectivity consistent with predictive coding, drawing on dynamical causal modeling studies of disorders of consciousness including vegetative states. The framework implies that anesthesia abolishes consciousness not by eliminating neural activity outright, but by disrupting the recurrent message-passing between hierarchical levels that is necessary for prediction error minimization. When top-down predictions cannot propagate to constrain lower-level error signals, or when precision weighting mechanisms are pharmacologically impaired, the system loses its capacity to select a dominant hypothesis with high posterior probability. Local neural activity may persist, but without the coordinated hierarchical inference that constitutes conscious perception, experience is abolished. Seth and Hohwy (2021) further note that PP furnishes resources for establishing mappings between biological mechanisms and phenomenological properties, and that disruptions of these hierarchical mechanisms in anesthesia, sleep, and seizures would all represent breakdowns in the conditions required for conscious inference.

For blindsight, the predictive coding framework offers an account based on the dissociation between residual processing and the hierarchical precision-weighted inference needed for consciousness. In normal vision, prediction errors generated in early visual cortex propagate through the cortical hierarchy, are assigned appropriate precision, and drive model updating that constitutes conscious visual experience. In blindsight, damage to V1 disrupts this normal pathway, but alternative subcortical routes may still generate prediction errors that influence behavior without achieving the precision weighting and hierarchical propagation required for conscious perception. Hohwy (2012) discusses how the model with the highest posterior probability --- spanning multiple levels of the cortical hierarchy --- determines conscious content. If visual information reaches motor or decision systems through pathways outside this hierarchical inference process, it can guide forced-choice responses without being selected for conscious experience. The split-brain case, while not directly addressed in these source papers, follows naturally from the framework: severing the corpus callosum would disrupt the flow of predictions and prediction errors between hemispheres, preventing the formation of a single unified generative model and plausibly resulting in two separate predictive systems each maintaining its own conscious contents.

PCT accounts for anesthesia abolishing consciousness while preserving neural activity through the distinction between the basic control hierarchy and the reorganization process that constitutes consciousness. The control hierarchy can operate automatically without consciousness -- its units compare perceptual signals against reference values and generate error-driven outputs as a purely mechanistic process. Consciousness, on the PCT account, is specifically tied to the reorganization process: the focused, error-driven modification of control system properties. Anesthesia would disrupt consciousness by interfering with the comparison and error-detection processes that trigger reorganization, or by blocking the propagation of error signals through the hierarchy to levels where they would recruit conscious attention. Even if basic neural signaling continues and some low-level control loops remain functional, the disruption of the capacity for reorganization-driven error monitoring would eliminate the conditions PCT identifies as constitutive of consciousness. Furthermore, Mansell (2024) notes that alterations in biological systems through drugs and medication directly impact intrinsic error states, affecting which perceptions are prioritized within consciousness. Anesthetic agents could suppress the intrinsic error signals that normally attract conscious awareness, leaving the organism in a state of automatic, non-conscious control.

Regarding split-brain cases, Mansell (2024) addresses dissociative phenomena by describing compartmentalization within the control hierarchy. In PCT, any multi-componential control system is at risk of functionally separate systems controlling variables that are also being controlled by other parts of the same system. Compartmentalization is adaptive for focused goal pursuit but can maintain conflict and entail loss of control in the form of dissociations. When the corpus callosum is severed, the integrated control hierarchy that normally coordinates reorganization across both hemispheres is fragmented into two largely independent hierarchical control systems. Each hemisphere would maintain its own reorganization process attracted to its own unresolved errors, potentially yielding two separate streams of consciousness. Mansell further proposes that continuity in consciousness may be regained only by information integration across compartmentalized systems through the development of a higher-level control system via reorganization -- a process that callosotomy would permanently prevent.

PCT offers a natural account of blindsight through the distinction between automatic control and conscious reorganization. Young (2026) notes that blindsight demonstrates that perception and behavior can occur without conscious experience, since individuals with this condition claim an inability to see objects within their visual field yet can still identify and interact with them when prompted. In PCT terms, visual processing in blindsight patients continues through control loops that handle errors automatically without engaging the hierarchical levels where consciousness monitors error and directs reorganization. The subcortical pathways that remain intact can support perceptual control of visually guided behavior, but because the primary visual cortex damage disrupts the pathways through which visual processing generates the kind of novel perceptual integration and reorganization-attracting error that constitutes conscious vision, the patient lacks visual experience despite preserved visual function. The patient's denial of seeing reflects the absence of quality perception and reorganization-driving error at conscious levels, while above-chance behavioral performance reflects intact lower-level automatic control.

Illusionism accounts for the loss of consciousness under anesthesia not as the elimination of phenomenal properties from the world, but as the disruption of the introspective mechanisms that generate the illusion of phenomenal consciousness. Sensory processing may continue at some level under anesthesia, but if the introspective systems that create quasi-phenomenal representations are suppressed, there will be no appearance of phenomenal experience and no subsequent report of having been conscious. Since on the illusionist view what makes experiences seem "like something" is introspective awareness -- representational mechanisms that monitor sensory states and generate the appearance of phenomenality -- the question of why anesthesia abolishes consciousness becomes the empirical question of which neural processes support this introspective capacity and how anesthetic agents disrupt them. This is a more tractable question than explaining why a special phenomenal quality vanishes, because it requires only standard neuroscientific explanation of a representational mechanism's failure.

For split-brain cases, the illusionist framework suggests that the apparent unity of consciousness is itself an aspect of the introspective illusion rather than a deep phenomenal fact. Frankish, drawing on Dennett, rejects the idea of a "Cartesian theatre" where unified experience is assembled for an inner observer. Instead, introspective representations may be generated by multiple subsystems distributed across the brain, with the work of appreciating and reacting to these representations parceled out to various subsystems. The illusion of unity would depend on integration among these introspective processes. When the corpus callosum is severed, each hemisphere may generate its own introspective representations independently, creating separate illusions of consciousness rather than one unified illusion. This is consistent with Dennett's multiple-drafts model, in which there need not be a single integrated stream of introspective representations; there might instead be numerous micropresentations yielding multiple drafts of sensation.

Blindsight receives a particularly natural explanation within the illusionist framework. In blindsight, visual processing continues and guides behavior, but the patient reports no visual experience in the affected field. For the illusionist, what is absent in blindsight is not some genuine phenomenal quality of seeing but rather the introspective representation that normally accompanies visual processing and creates the illusion of visual phenomenality. The visual information is processed and can influence forced-choice responses, but because it does not become the target of the introspective system that generates quasi-phenomenal representations, the patient sincerely and accurately reports not seeing. Normal vision involves visual processing plus introspective representation that creates the appearance of phenomenal seeing; blindsight involves visual processing without that introspective component. The dissociation thus directly illustrates the illusionist's distinction between the underlying sensory processing and the introspective overlay that creates the illusion of phenomenal consciousness.

Orch OR offers a specific and distinctive account of anesthesia. The theory proposes that anesthetic gases --- despite their extraordinary molecular diversity --- converge on a common mechanism: binding within hydrophobic pockets of proteins, including and especially tubulin in microtubules. Within these hydrophobic regions, van der Waals London force interactions among non-polar electron clouds govern the dipole states that Orch OR identifies as quantum bits (qubits) underlying consciousness. Anesthetics, by occupying these hydrophobic channels, disperse the collective London force dipole couplings that are necessary for quantum superposition and coherent dipole oscillations along microtubule lattice pathways. Emerson et al. demonstrated that cessation of tadpole behavior occurs specifically via anthracene anesthetic binding in microtubules, and genomic and proteomic evidence points to microtubules as a primary site of anesthetic action. On this account, anesthesia abolishes consciousness while preserving classical neural signaling because it selectively disrupts the quantum-level processes in microtubules --- the dipole superpositions and their orchestrated reduction --- without disabling the conventional electrochemical machinery of neuronal firing. Neurons continue to transmit action potentials and process information classically, but the quantum coherence required for Orch OR is eliminated, and with it, conscious experience.

Regarding split-brain cases, the source papers do not provide an extended treatment, but the theoretical framework yields clear implications. Orch OR proposes that quantum states in microtubules of one neuron can extend to microtubules in adjacent neurons via gap junctions --- primitive electrical connections that fuse cytoplasmic interiors into a continuous "hyper-neuron" or syncytium. When gap junctions are open, quantum entanglement can spread across large populations of neurons, enabling the brain-wide orchestration of microtubule quantum states that the theory identifies with unified conscious experience. Severing the corpus callosum would disrupt callosal gap junction pathways (and conventional axonal connections) between hemispheres, preventing quantum entanglement from extending across the midline. Each hemisphere's microtubule networks would then undergo independent Orch OR events, plausibly resulting in two separate streams of consciousness --- consistent with the clinical observation that split-brain patients exhibit dual centers of awareness. The binding of consciousness into a unified experience depends, in this framework, on the physical continuity of quantum coherence across the brain, which callosal section would interrupt.

For blindsight, Orch OR suggests that visual information reaching the brain through subcortical pathways after damage to primary visual cortex can still be processed classically by neurons whose microtubules operate in non-quantum modes, enabling accurate forced-choice responses without conscious visual experience. The critical element is that the damaged cortical region can no longer support the orchestrated quantum superpositions and subsequent OR events that would constitute phenomenal seeing. The subcortical route delivers information that influences behavior through classical microtubule automata and conventional neural processing, but this processing never engages the quantum-gravitational mechanism of objective reduction. The patient thus has access to visual information --- they can guess correctly about stimuli --- but lacks the conscious visual experience that would only arise if the relevant cortical microtubule networks were intact and capable of sustaining orchestrated quantum coherence to the point of self-collapse.

DLCT does not address these specific clinical phenomena in detail within the source papers, but its framework provides principled explanations for each dissociation. Regarding anesthesia, the theory would predict that anesthetic agents disrupt the self-referential feedback control mechanism that connects macro-level psychological laws to micro-level neural activity. Since consciousness in DLCT depends on the active interplay between two causal levels --- macro-level algebraic structural constraints driving micro-level synaptic adjustments --- anesthesia could abolish consciousness by interrupting this feedback loop while leaving basic neural processing at the micro level intact. The micro-level neural circuits (circuit A) responsible for sensory processing and homeostasis would continue to function under micro-level neural laws, but the macro-level psychological laws could no longer exert downward causation on neural states, eliminating the dual-level dynamic that the theory identifies as essential for consciousness.

For split-brain cases, DLCT's emphasis on modular structure and inter-module algebraic constraints provides a natural account. The theory proposes that macro-level feedback errors are defined by the relationships among multiple neural network modules --- for example, commutativity constraints between pairs of modules. Severing the corpus callosum would physically disconnect the modules across hemispheres, making it impossible to compute or enforce unified macro-level algebraic constraints across the full set of modules. Each hemisphere could potentially develop its own self-referential feedback control mechanism with its own set of macro-level psychological laws operating over its available modules, resulting in two separate streams of consciousness. The unity of consciousness, on this account, depends on the physical connectivity necessary to maintain coherent algebraic structural constraints across modules, and disrupting that connectivity fragments the macro-level causal structure.

For blindsight, the theory would predict that visual information processed through residual subcortical pathways engages only micro-level neural circuits (circuit A) without entering the self-referential feedback control mechanism. Normal conscious vision would require that visual processing participates in the dual-level dynamics where macro-level psychological laws constrain and are informed by visual representations. In blindsight, the relevant pathways to the macro-level feedback system are damaged, so visual information influences behavior through micro-level processing alone --- sufficient for forced-choice discrimination but lacking the macro-level intrinsic causal power that the theory associates with conscious experience. The patient perceives nothing because the processing never generates the macro-level feedback errors whose reduction, through downward causation, constitutes the mechanism of conscious perception.

Irruption theory's account of these dissociations follows from its core distinction between irruption and absorption as separable material signatures of mind-matter interaction. In the case of anesthesia, the theory would predict that what is disrupted is specifically the absorption dynamic -- the capacity of neural processes to compress into the low-dimensional structures that correlate with conscious experience. Neural activity persists because the basic physical and metabolic operations of neurons remain intact, but the specific transformation by which material activity makes a difference to the mind is blocked. Froese notes that methodological proxies for entropy production, such as broken detailed balance, irreversibility, and state diversity, are being developed to track levels of conscious awareness, and the theory would predict that anesthesia specifically disrupts these markers. Importantly, irruption theory also predicts that anesthesia should reduce the irruption signature -- the entropy-increasing, unintelligible variability associated with agential efficacy -- because if the mind is no longer being affected by material processes (loss of absorption), then the mind's capacity to affect material processes (irruption) should correspondingly diminish. The bidirectional coupling through the black-box middle would be suppressed in both directions.

For split-brain cases, irruption theory aligns with the general prediction that any material system embodying a conscious agent must be a composite system, and it emphasizes that the bidirectional mind-matter relation has spatiotemporal structure. The theory's discussion of biological organization predicts distinct physiological categories related to irruption-dominant and absorption-dominant processes, and these are expected not to directly overlap in space. When the corpus callosum is severed, the spatiotemporal coordination that normally enables unified absorption across hemispheres would be disrupted, potentially yielding independent absorption processes and thus partially independent streams of consciousness. The theory's analysis of traveling waves in the brain's functional systems -- interpreted as providing an optimal solution to the incompatible spatiotemporal markers of irruption and absorption -- suggests that unified consciousness depends on the large-scale integration that such waves provide. Severing the major commissure would fragment this integration.

Blindsight poses a particularly interesting case for irruption theory. The theory would account for it by arguing that visual processing in blindsight patients retains its capacity to influence behavior (through subcortical pathways) without undergoing absorption -- without the specific compression and dimensionality-reduction that would make that processing contribute to conscious visual experience. The visual information drives motor and behavioral responses but never passes through the black-box interface into the mind. This is consistent with the theory's prediction that action and perception have contrary material signatures and can be dissociated: blindsight preserves the capacity for visually guided action (which involves irruption-like diversification of motor possibilities) while losing the conscious perceptual experience (which requires absorption-like compression). The theory would predict that neural measures in blindsight patients should show intact processing complexity in visual pathways without the specific compression and redundancy signatures that characterize conscious visual awareness.

The FEP-based model accounts for anesthesia's abolition of consciousness through the disruption of the neuromodulatory systems that constitute the active states of the innermost Markov blanket. On this account, consciousness depends not on neural activity per se but on the specific capacity for covert action --- the precision-weighting and attentional gating mediated by ascending modulatory neurotransmitter systems. Friston notes that altered states of consciousness, including those induced by anesthesia and sleep, "speak directly to the enactive aspects of inference," and that the hallmark of reduced levels of consciousness is an absence of responsiveness. During anesthesia, the pharmacological agents target precisely the neuromodulatory systems responsible for controlling the sparse coupling among neuronal populations --- the same systems that the theory identifies as implementing covert action on the inner screen. When these systems are disrupted, the generative model loses its "temporal thickness" and becomes as thin as that of a virus: the system can still produce reflexive responses that minimize surprise in the "here and now," but it can no longer entertain counterfactual hypotheses about the future, plan actions, or deploy attention. The generative model, in Friston's formulation, loses its depth, and with it the capacity for consciousness.

Regarding split-brain cases, the FEP framework provides a natural account through the hierarchical assembly of Markov blankets. The theory holds that when components of a joint system are non-mutually-isolated and share a boundary, an internal Markov blanket mediates classical communication between them, serving as an internal classical memory structure. When this internal boundary is severed --- as in callosotomy --- the composite system loses the internal Markov blanket that previously unified the two hemispheres into a single nested hierarchy. According to the hierarchical assembly process described in the theory, when two components are added to a multi-component system, the joint system's environment decreases and its Markov blanket reconfigures. Severing the callosal connection would effectively decompose what was one irreducible Markov blanket structure into two separate hierarchies, each with its own innermost screen and its own associated conscious experience, consistent with the clinical observation of two apparently independent streams of consciousness in split-brain patients.

The FEP model can address blindsight through the distinction between overt and covert action within the nested hierarchy. In blindsight, patients retain the ability to make accurate behavioral responses to visual stimuli they deny seeing. The theory would attribute this to the preservation of overt action pathways --- the lowest levels of the hierarchy where active states couple directly to neuromuscular junctions via classical reflexes --- while the covert action pathways that enable conscious visual experience are disrupted. Since the theory identifies consciousness specifically with the information selected by covert action (precision-weighting via neuromodulatory systems), damage to the cortical pathways that participate in this attentional selection would eliminate conscious visual experience while leaving intact the subcortical and lower-level cortical circuits that can still drive overt motor responses. The visual information is still being processed through the Markov blanket hierarchy, but it is not being "read" by the innermost screen through the mechanism of covert mental action that the theory identifies as necessary for conscious perception.

Recurrent Processing Theory offers a unified framework for understanding these dissociations by proposing that each involves a disruption of recurrent processing while leaving feedforward processing relatively intact. For anesthesia, Lamme draws on the observation that many anesthetic agents have as their final common pathway the blockade of NMDA receptor activation. He cites evidence from his own laboratory demonstrating that figure-ground activity in primary visual cortex -- a signature of recurrent processing -- is suppressed by anesthesia, while feedforward signals remain relatively untouched (Lamme et al., 1998b). In the 2010 paper, he further notes that blockade of NMDA receptors in monkey visual cortex reduces recurrent signals, while blockade of AMPA receptors has its main effect on feedforward activity. Since RPT identifies recurrent processing as the neural basis of consciousness, the selective disruption of NMDA-mediated recurrent interactions by anesthetic agents explains why consciousness is abolished even though substantial neural activity persists. The brain continues to process sensory information in a feedforward manner under anesthesia, but without the recurrent loops that constitute phenomenal experience, there is nothing it is like to be the anesthetized subject.

For blindsight, the theory provides a particularly clean account. Damage to V1 eliminates the critical hub through which feedback signals from higher visual areas return to early visual cortex, thereby destroying the recurrent processing loops that normally support visual consciousness. Visual information can still reach higher areas such as MT and even ventral stream regions through alternative subcortical or extrastriate pathways, enabling residual feedforward processing that supports above-chance guessing about stimulus properties, localization, and even emotional responses. But without recurrent interactions involving V1, these residual capacities remain unconscious. This is consistent with TMS evidence showing that disruption of V1 at the time when feedback signals would normally arrive (around 100 ms post-stimulus) eliminates visual awareness while preserving some unconscious processing, and with findings that in monkeys with V1 lesions, ventral stream areas no longer respond to stimuli while parietal cortex (e.g., area MT) still does. The theory thus explains the characteristic dissociation of blindsight: patients cannot detect stimuli (no conscious experience because no recurrent processing), yet can perform above chance on forced-choice tasks (residual feedforward processing supports unconscious behavior).

For split-brain cases, Lamme's framework offers an especially provocative interpretation. Rather than concluding that the right hemisphere is unconscious because it cannot support verbal report, RPT argues that the split-brain case is better understood as a manipulation of language and reportability rather than a manipulation of consciousness itself. The right hemisphere still possesses the cortical architecture necessary for recurrent processing within its own visual areas. Patients can draw objects presented to the left visual field, select them from alternatives, and perform cognitive operations with the left hand -- behaviors that suggest recurrent processing and hence, on the RPT account, phenomenal consciousness within the right hemisphere. Lamme argues that the failure to verbally report these experiences reflects a disruption of the access mechanisms (Stage 4 processing involving language areas) rather than an absence of phenomenal consciousness (Stage 3 recurrent processing). The theory would predict that each hemisphere sustains its own independent domain of phenomenal experience through localized recurrent processing, and that what is lost when the corpus callosum is severed is the ability to integrate these domains into a single unified access-conscious experience, not the phenomenality within each hemisphere.

AST's account of blindsight is among its most developed clinical applications. Webb and Graziano (2015) describe blindsight as a condition in which patients lack subjective awareness of stimuli in part of their visual field while retaining above-chance objective discrimination of those stimuli. In AST, this dissociation has a natural explanation: the visual information is still being processed and can still influence behavior through attentional mechanisms, but the attention schema has become disconnected from or fails to model the attentional state directed at the affected region. The patient's brain processes the stimulus and attention is drawn to it, but the internal model of attention --- the attention schema --- does not register this attentional engagement. As a result, the patient has no basis for reporting awareness of the stimulus, even though the stimulus is being processed and attention is, in some residual sense, allocated to it. Webb and Graziano (2015) extend this logic to hemispatial neglect, in which patients lose subjective awareness of stimuli on one side of space while sometimes retaining unconscious processing of those stimuli. Graziano (2022) identifies the temporoparietal junction, particularly the right TPJ, as a critical brain area for constructing the attention schema. Damage to this region produces the most severe forms of neglect, involving both a loss of conscious experience and a loss of endogenous attentional control on the affected side, consistent with AST's prediction that disrupting the attention schema should impair both awareness and the top-down control of attention simultaneously.

Regarding anesthesia, AST does not provide an extensively developed mechanistic account of how specific anesthetic agents abolish consciousness while preserving neural activity. However, the framework implies a clear logic: if consciousness depends on the brain's capacity to construct and maintain an attention schema --- a model of its own attentional processes --- then any pharmacological or physiological intervention that disrupts the neural systems responsible for computing that model would abolish awareness while potentially leaving lower-level sensory and attentional processing partially intact. The dissociation between preserved neural activity and absent consciousness is exactly what AST would predict if anesthesia selectively disrupts the higher-order modeling process rather than all neural processing. For split-brain cases, AST's framework similarly implies that if the attention schema depends on integrating information across hemispheres to construct a unified model of attention, then severing the corpus callosum could produce two separate attention schemas, each modeling only the attentional processes of its respective hemisphere. While AST does not elaborate a detailed mechanistic prediction for the split-brain case in these papers, the logic of the framework accommodates the observation: a unified consciousness requires a unified attention schema, and physically dividing the substrate that computes the schema should divide the resulting consciousness.

NPS does not provide detailed accounts of anesthesia, split-brain phenomena, or blindsight in the source papers. Because the theory focuses on the specific content of consciousness rather than on its generic presence or absence, the question of why anesthesia abolishes consciousness while preserving neural activity falls largely outside its current explanatory scope. However, the framework does provide resources that bear on dissociation phenomena in a general way. The central insight would be that what matters for conscious experience is not the mere presence of neural activity but the preservation of the relevant relational structures --- the Q-N structural mirroring. If anesthesia disrupts the structural organization of neural activation spaces such that they no longer mirror the corresponding phenomenal quality spaces (even while substantial neural activity persists), then consciousness would be abolished on NPS's terms. The theory's emphasis on structure over raw activity is consistent with the observation that neural firing alone is insufficient for consciousness.

For split-brain cases, NPS's commitment to neurophenomenal holism is relevant. The theory holds that any single phenomenal experience is fully individuated by its place in the whole Q-structure, and any change in the web of phenomenal relations can alter the character of every experience within it. Severing the corpus callosum would fragment the holistic neural structure that mirrors the unified phenomenal space, producing two separate Q-structures mirrored by two separate N-structures. For blindsight, NPS could in principle argue that the residual visual processing in blindsight patients fails to instantiate the relevant structural similarity with the full phenomenal quality space for vision --- the neural activation patterns do not form a proper surjective homomorphism onto the visual Q-structure --- which is why there is visual function without visual phenomenal experience. That said, the source papers do not develop these accounts in any detail, and NPS explicitly leaves the mechanisms of generic consciousness to other theories, positioning itself as a constraint on identifying the specific neural substrates of conscious content rather than as a theory of when consciousness is present or absent.

Regarding anesthesia and states in which consciousness is abolished while neural activity persists, the embodied cognition framework offers a principled account grounded in the disruption of sensorimotor engagement. On the sensorimotor contingency view, consciousness does not consist in any particular state of neural activation, but in the organism's active exercise of mastery over sensorimotor contingencies, integrated with capacities for thought and action-guidance. Anesthesia abolishes consciousness not primarily by silencing neural activity, but by disrupting the organism's capacity to actively explore its environment and exercise the sensorimotor skills that constitute perceptual experience. When the brain can no longer sustain the dynamic, law-governed coupling between the organism's actions and the resulting sensory changes --- when it can no longer support the knowledge-mediated exploration that defines seeing, hearing, or touching --- consciousness ceases, even though substantial neural activity may continue. Varela's neurophenomenological perspective would add that the neural synchrony and large-scale integration characteristic of conscious states collapse under anesthesia, severing the reciprocal circulation between embodied action and experiential awareness that his method demands.

For blindsight, O'Regan and Noe provide a detailed account. Patients with lesions in the visual cortex who can make accurate forced-choice judgments about stimuli in their scotoma are, on their view, not genuinely seeing. These patients are sensitive to stimuli --- information is being processed and can influence behavior --- but they lack visual consciousness of the stimuli because they do not actively probe those stimuli through the exercise of sensorimotor contingencies integrated with thought and action-guidance. The theory holds that one does not hear the sound of a drill before noticing it, and one does not see a stimulus in the blind field before attending to it: prior to the active probe, prior to the exercise of the relevant sensorimotor mastery in the service of thought and reflection, there is no perception. The empirical data on change blindness and inattentional blindness strongly support this view --- observers can be looking directly at large, fully visible changes and still not see them, because seeing requires not just retinal stimulation but active attentional engagement governed by sensorimotor knowledge.

The theory does not address split-brain cases in extensive detail, but its framework provides conceptual resources for approaching them. Since visual awareness, on this account, consists in the organism's integrated exercise of sensorimotor mastery across its full behavioral repertoire --- including speech, reasoning, and action planning --- the severing of callosal connections would be expected to disrupt the integration of sensorimotor exploration across the two hemispheres. Each hemisphere might retain its own domain of sensorimotor mastery, but the unified consciousness that depends on integrating these capacities into a single coherent agent's thought and action-guidance would fragment. The two-visual-systems framework (dorsal and ventral streams) that O'Regan and Noe discuss at length is relevant here: visual awareness depends on a broad range of capacities for bodily guidance, speech, and rational thought, and to the extent that these capacities are dissociated by callosal section, corresponding dissociations in conscious experience would be expected.

DIT provides its most detailed and compelling account for the case of anesthesia. The theory explains anesthetic-induced unconsciousness not as a wholesale shutdown of neural activity, but as a selective decoupling of the apical and somatic compartments of L5p neurons. Virtually all known general anesthetics, despite their diverse molecular mechanisms, share a common target in the non-specific thalamus. According to DIT, anesthetics disrupt the thalamic modulation that normally enables apical-to-somatic signal propagation within L5p cells. This was directly demonstrated by Suzuki and Larkum (2020): under anesthesia, optogenetic stimulation of the apical dendrites of L5p neurons failed to influence somatic firing, whereas the same stimulation in awake animals produced robust effects. Furthermore, Murayama and Larkum (2009) showed that apical dendritic calcium activity in L5p cells was four-fold weaker in the anesthetized state compared to quiet wakefulness. Critically, primary sensory pathways (the specific thalamic pathways projecting to layer 4) remain functional under anesthesia, which is why the brain continues to show evoked responses to stimuli. What is abolished is the integration of this feedforward information with contextual modulation via the non-specific thalamic loop -- precisely the mechanism DIT identifies as necessary for consciousness.

Regarding blindsight and split-brain phenomena, DIT offers a framework but with less direct experimental support. For blindsight, the theory would predict that visual information processed through intact feedforward pathways (bypassing the damaged primary visual cortex) can still drive behavior through cortical processing that does not adequately engage L5p neurons or the thalamo-cortical broadcasting loop. The visual information reaches cortical areas and influences responses, but because it is not integrated into the L5p-mediated thalamo-cortical system, it remains unconscious. For split-brain cases, DIT's emphasis on cortico-cortical loops as essential for computing conscious contents suggests that severing the corpus callosum would disrupt the long-range cortico-cortical connectivity needed for unified integration at L5p neurons across hemispheres. However, DIT also proposes that the thalamus, which remains intact in split-brain patients, could partially maintain some integration, as the non-specific thalamic nuclei project diffusely across the cortex. The theory acknowledges that the thalamic projection patterns are necessarily simplified in its current formulation, and the detailed application to split-brain phenomenology remains an area requiring further development.