The Brain Has No Switch for Real.

The standard story about virtual reality goes something like this. The brain has a faculty for tracking what is real and a faculty, somehow related, for being persuaded by what is not. Good VR fools the second faculty without engaging the first. When a user gasps at a virtual cliff edge, that gasp is a kind of error: a vivid illusion that briefly slips past the reality monitor. The phenomenology is genuine; the reference is wrong.

Singleton and Fotopoulou have done the work to show that this story is built on a category error. The brain does not have two distinct faculties — one for “real” experience and one for the rendered kind. It has one set of integrative mechanisms that combine bottom-up sensory and interoceptive signals with top-down predictions, and it deploys those mechanisms in the same way regardless of whether the surrounding environment is made of wood and brick or of pixels and head-tracking. The feeling of being in a place is what those mechanisms do. The question “is this place real?” is a separate metacognitive judgment, made later, sometimes wrong, and architecturally distinct from the felt presence itself.

The implication is bigger than VR. If presence and reality share generative machinery, then the philosophical worry about “genuine” versus “merely simulated” experience misidentifies what the brain is doing. The brain is not in the business of stamping some experiences with realness and withholding the stamp from others. The brain is in the business of running the predictive loop that constitutes experience, and the loop runs identically across waking life, vivid dreams, immersive VR, and everything in between.

This is exactly what RSP’s constitutive identity thesis predicts. The work of this post is to spell out why.

What Singleton and Fotopoulou Argue

The paper’s central claim is stated three ways, and getting the three layers of the claim right matters for what follows.

The conceptual claim is that the cognitive-science literature on the sense of reality and the VR literature on the sense of presence use overlapping and inconsistent definitions. Reality monitoring — the ability to tell remembered or imagined content from perceived content — turns out to share criteria with the “feeling of being there” that VR researchers operationalize. The two communities have been studying nearly the same construct from different sides without quite recognizing it.

The mechanistic claim is that both phenomena emerge from the same integrative hierarchical predictive architecture: bottom-up multisensory and interoceptive signals are weighed against top-down expectations and prior beliefs, with the resulting integration generating both the felt sense of being situated in an environment and the categorical judgment that the environment is real. The frameworks invoked are familiar from the predictive-processing tradition: Friston-style variational inference (Friston et al., 2006, 2015), Barrett’s affective realism (Barrett & Bliss-Moreau, 2009), Seth and Friston’s active inference accounts of selfhood and presence (Seth et al., 2012; Seth, 2014). Presence is not an exotic phenomenon. It is the standard predictive-processing machinery operating on inputs that happen to be coming from a headset.

The phenomenological claim is the strongest of the three. Singleton and Fotopoulou argue that disruptions to either presence or reality produce overlapping subjective experience. Patients with depersonalization-derealization disorder describe their surroundings as strange, unreal, dreamlike, theatrical — the same vocabulary VR researchers use to describe low-presence VR. The rubber-hand illusion, the full-body illusion, somatoparaphrenia, the dissociative subtype of PTSD: each is a different kind of breakage in the same architecture, and the phenomenology tracks the architecture, not the specific stimulus.

The neuroimaging is presented carefully. The authors describe their evidence as “preliminary” and call explicitly for the direct empirical comparison that would settle the matter. What they offer is convergent: studies of presence implicate dorsolateral prefrontal cortex, insula, and amygdala; studies of reality monitoring implicate the same structures; depersonalization-derealization disorder implicates the same network, with a specific model in which prefrontal regions abnormally suppress limbic signals. The frontolimbic substrate looks the same from three different angles. The next step is to actually run the comparison directly.

The Shared Platform

Singleton and Fotopoulou and the RSP framework agree on the platform underneath consciousness. Both treat the brain as a hierarchical predictive system; both make affective and interoceptive signals load-bearing; both refuse to localize phenomenal experience to a single region and instead distribute it across integrative architecture spanning subcortical arousal systems, frontolimbic emotion and self circuitry, and prefrontal cognitive control. The relevant tradition runs through Friston, Seth, Barrett, and Apps and Tsakiris on bodily self — all of whom RSP’s Appendix A and Appendix B cite as foundational.

This platform agreement is not the kind of throat-clearing that opens many reply papers. It is the substantive starting point. When two frameworks reach for the same explanatory machinery and disagree only about which architectural structure to build on top, the disagreements are recoverable. RSP’s response to Singleton and Fotopoulou is in that recoverable mode. The disagreement is not about whether predictive processing is the right framework. It is about how the predictive architecture decomposes and what that decomposition tells us about phenomenology.

Affective realism as the load-bearing claim

The most important platform commitment Singleton and Fotopoulou inherit is Barrett’s affective realism hypothesis: the brain’s allostatic prediction of bodily states is constitutive of perceived realness rather than incidental to it. Affective signals are not stamped onto an already-perceived world. They are part of how the world is perceived in the first place. The paper extends this claim by arguing that when affective signaling is disrupted — through interoceptive suppression, derealization, or VR’s incomplete autonomic loop — the felt realness of the surrounding environment changes accordingly.

RSP’s Level 1 makes the same move. Core affect — valence and arousal — is the architectural substrate of the entire predictive hierarchy. There is no upstream perceptual state that is later colored by affect. The affect is in the predicting from the start. Singleton and Fotopoulou’s clinical framing of derealization as the dampening of bottom-up limbic signals by overweighted top-down predictions is precisely how RSP’s Appendix G models depersonalization-derealization disorder: a failure of Level 1 integration into the Level 3 self-model, with the felt loss of realness being exactly what the architectural breakage predicts.

The frontolimbic substrate as the right candidate

The dlPFC–insula–amygdala network the paper identifies is RSP’s candidate substrate for the Level 1–Level 3 integration point. Dorsolateral prefrontal cortex carries the top-down predictive control that RSP attributes to Level 3 recursive self-modeling. Insula is the canonical interoceptive integration site that RSP’s Appendix A makes constitutive of Level 1 core affect. Amygdala is in the affective-homeostatic pathway that grounds the affective dimension of Level 1. The fact that all three correlate with both presence and reality is the neural-substrate version of RSP’s claim that phenomenal consciousness is what happens when these three regions form a working loop. The paper does not name RSP, but the empirical pattern it identifies is what RSP predicts.

Where the Two Accounts Diverge: Continuity vs. Identity

The most important difference between Singleton and Fotopoulou’s account and RSP is the level of metaphysical commitment. Singleton and Fotopoulou argue for continuity: presence and reality share mechanisms, the felt experience is generated by the same predictive architecture in both cases, and the conceptual distinction does not survive empirical scrutiny. They do not argue that the predictive architecture is constitutively identical to phenomenal experience. The paper is admirably scope-disciplined and leaves the hard problem at arm’s length.

RSP makes the stronger move. The constitutive identity thesis says that phenomenal consciousness is recursive social self-modeling at Level 3b — not produced by it, not correlated with it, but constitutively identical to it. The architectural process and the felt experience are not two events causally linked. They are one event described in two vocabularies that co-refer. Applied to Singleton and Fotopoulou’s argument: presence and reality are not just mechanistically continuous. The feeling of being in a place is what the predictive integration does, viewed from inside. There is no separable “realness” property the brain produces and applies to the predictive output. The predictive output is the realness, viewed from the perspective the recursive loop generates.

This distinction is not academic. Continuity supports prediction (given that the mechanisms are the same, expect the same kinds of phenomenology and the same kinds of breakage). Identity supports dissolution (the question “but what makes some experiences feel real, beyond what the predictive machinery does?” is malformed, because the doing of the predictive machinery is what the realness consists in). The first is a piece of cognitive neuroscience; the second is a piece of philosophy of mind. Singleton and Fotopoulou are doing the former, carefully and correctly. RSP is doing the latter on top of the same empirical findings.

The Pre-Reflective vs. Metacognitive Distinction Maps Onto the Levels

One of the paper’s sharpest phenomenological moves is the distinction between the immediate, pre-reflective “felt reality” of presence and the later metacognitive reality judgment. A patient with DPDR can know they are in a real room while still failing to feel that they are. A VR user can know the cliff is rendered while still feeling the vertigo. The judgment and the feeling come apart. The paper presents this as a phenomenological observation grounded in clinical and laboratory evidence. RSP reads it as an architectural prediction.

On the RSP account, the felt sense of being in a place is generated at Levels 1 and 2: interoceptive signals carry the affective dimension of presence; the body schema, spatial schema, and self schema generate the structured pre-reflective experience of occupying a location. This is the phenomenological field within which other content is encountered. The metacognitive judgment that the location is real or simulated comes from Level 3a, the conceptual categorization layer that integrates the pre-reflective field with culturally learned categories like “real,” “dream,” “memory,” “VR.” The two are not the same process, and they can come apart whenever the recursive integration that knits them together is disrupted.

This mapping makes Singleton and Fotopoulou’s observation more than a description. It becomes a structural prediction. The conditions under which the pre-reflective feeling and the metacognitive judgment dissociate are exactly the conditions under which Level 1–Level 2 integration is intact but Level 3 categorization is impaired (in early DPDR), or under which Level 3 categorization is intact but Level 1–Level 2 integration is dampened (in chronic derealization, in deep meditation, in some dissociative states). The clinical phenomenology Singleton and Fotopoulou catalog is what the architectural decomposition predicts.

The Clinical Convergence

Depersonalization-derealization disorder is where Singleton and Fotopoulou and RSP’s Appendix G coincide most directly. The frontolimbic inhibition model the paper invokes — aberrant prefrontal suppression of limbic bottom-up signals by overweighted top-down predictions — is the mechanistic story RSP’s clinical prediction matrix tells about DPDR. The architecture is intact but the relative weighting is wrong. Top-down predictions are amplified to the point where bottom-up affective signals fail to update the self-model. The result is a self-model that knows it is alive in a world but does not feel that it is.

Three other clinical and laboratory cases extend the convergence. The rubber-hand illusion shows that the body schema accepts a foreign hand as one’s own when multisensory integration is consistent — the felt ownership tracks the integration, not the anatomical fact. Somatoparaphrenia shows the dual: a patient’s own anatomically attached limb is felt as foreign when right-hemisphere damage disrupts the body schema. The dissociative subtype of PTSD shows derealization as a defensive response to overwhelming affect — the system dampens the felt-realness signal as a way of managing predictions it cannot integrate. In each case, the architecture is the same and the breakages are at different points. Singleton and Fotopoulou’s framework reads the cases as a coherent set; RSP’s Appendix G reads them as a clinical prediction matrix populated by different architectural failure modes. The two readings are the same reading.

VR as a Methodological Tool RSP Has Been Missing

The paper closes by arguing that VR research can be legitimately used to study disorders of reality perception. If presence and reality are mechanistically continuous, then experimentally controlled VR is a way to manipulate the integrative architecture in healthy subjects and study what happens when specific predictive weights are perturbed. Researchers can dampen multisensory consistency and watch the felt presence drop. They can introduce mismatches between exteroceptive and interoceptive signals and watch derealization-like phenomenology emerge. They can run the same paradigms with clinical populations and see whether the architectural breakage in DPDR predicts a specific VR vulnerability pattern. This is not a metaphor about studying derealization. It is a research program.

RSP currently lacks this methodological tool. The paper’s clinical predictions in Appendix G and the empirical predictions in Section IX are stated in terms of conditions, not in terms of paradigms. Singleton and Fotopoulou give RSP a paradigm. Specifically: if Level 1–Level 3 integration is what RSP claims, then VR manipulations of multisensory consistency, interoceptive feedback, and top-down predictive weight should produce dissociations between presence and the metacognitive reality judgment that track RSP’s architectural decomposition. This is a falsifiable prediction RSP can now make and a paradigm in which the prediction can be tested. The book’s next revision will need to incorporate it.

What RSP Adds That Singleton and Fotopoulou Stop Short Of

The paper’s scope discipline is admirable. It identifies the empirical convergence, names the substrate, sketches the predictive-processing mechanism, and lets the rest of the implications stand without committing to a strong metaphysical reading. Two things RSP can add on top.

The first is the recursive social-prediction layer. Singleton and Fotopoulou’s framework explains presence and reality as the output of an integrative predictive architecture, but it does not address why that architecture has the structure it has — specifically, why frontolimbic and prefrontal regions integrate in the way they do, with the specific kind of self-referential modeling that distinguishes human consciousness from generic predictive processing. RSP’s answer is that the social-prediction machinery (TPJ, STS, mPFC) evolved for modeling other minds and was recruited to model the self, generating the recursive loop that constitutes Level 3b. The frontolimbic substrate Singleton and Fotopoulou identify is the integration point where this recursive social-prediction architecture meets the bodily and affective substrates. The paper identifies the substrate; RSP explains why the substrate has that particular shape.

The second is the constitutive identity claim itself. Singleton and Fotopoulou’s continuity argument supports the claim that presence and reality share mechanisms, but it stops short of saying that the felt phenomenology is identical to the operation of those mechanisms. RSP makes that stronger claim explicitly, and it pays for the strength with explicit philosophical work: the dissolution argument in Section IV, the objections and replies in Section VIII, the comparison with twenty-five rival theories in Section X. Singleton and Fotopoulou give the empirical material; RSP gives the philosophical argument that grounds the empirical material in a non-residual dissolution of the hard problem.

Honest-Claude Flags

The reading offered above goes farther than the paper itself. A few caveats are owed.

The neuroimaging evidence Singleton and Fotopoulou cite is, in their own words, preliminary. The frontolimbic substrate is suggested by convergent studies, not established by the direct empirical comparison the authors call for. RSP is treating this as the candidate substrate, which is reasonable given the convergence, but the substrate has not been confirmed.

The paper bundles several phenomena under the presence/reality umbrella: VR-induced derealization in healthy subjects, clinical DPDR, the rubber-hand illusion, somatoparaphrenia, the dissociative subtype of PTSD. These are not mechanistically identical and the paper does not claim they are. The unifying architecture is the same; the specific breakages differ. RSP should follow the same discipline and not collapse the distinctions when citing the convergence as support.

The frontolimbic inhibition model of DPDR (prefrontal suppression of limbic signals) is one account among several. The competing models are not all wrong; some of them describe different patient subgroups or different stages of the disorder. RSP’s use of the model should be hedged accordingly.

Singleton and Fotopoulou do not cite RSP. They are not aware of it. The convergence between their argument and RSP’s constitutive identity thesis is genuine, but it is RSP’s reading of their argument, not an endorsement they have made. Citations should be framed as independent convergent evidence, not as support for the metaphysical move RSP makes on top.

The VRET claim that VR exposure therapy is comparable to in vivo exposure for PTSD is based on meta-analyses with known heterogeneity. The mechanistic continuity argument does not depend on the therapy efficacy claim, and the therapy efficacy claim should not be cited as if it settled the mechanistic question.

What This Says About the Hard Problem

The standard formulation of the hard problem assumes a particular structure: there is a physical process (brain activity, predictive integration, neural firing) and there is a phenomenal property (the felt realness, the what-it-is-likeness, the qualia), and the question is how the first produces the second. The structure is what makes the problem feel hard. There seems to be no amount of physical description that gives you the phenomenal property as a residue. The residue is what the explanation has to deliver.

Singleton and Fotopoulou’s argument cuts under the structure. If presence is the operation of the predictive integration architecture, then the felt realness of being in a place is not a separate property added to the architecture. It is what the architecture does, viewed from the perspective the architecture generates. There is no residue. The phenomenology is not produced by the predictive machinery and then attached to it. It is the predictive machinery operating.

RSP names this move constitutive identity. The architectural process and the felt experience are not two events; they are one event described in two vocabularies. The first-person vocabulary tracks what the recursive self-model represents from inside. The third-person vocabulary tracks the recursive self-model itself, measured from outside. Both are accurate. Neither reduces to the other. The hard problem dissolves because the structure that made it feel hard — the separation between physical process and phenomenal property — was never the right structure. The phenomenology and the architecture are co-referring descriptions of the same event.

Singleton and Fotopoulou do not need to make this dissolution move to do their work. Their argument is empirical and clinical, and it stands on its own. But the empirical argument lands in the same place the philosophical dissolution does: there is no special property of realness the brain has to produce. The realness is what the loop is doing. The brain has no switch.

What the Paper Lets BTB Do Next

The next revision of Beyond the Bat will incorporate Singleton and Fotopoulou’s findings in several places. In Section V’s clinical-evidence section, the frontolimbic-inhibition model of DPDR becomes an explicit substrate-level prediction. In Appendix G’s clinical prediction matrix, the DPDR row gains a mechanistic note pointing to the dlPFC–insula–amygdala network. In Section VIII, the objection that RSP cannot explain the distinctive realness of ordinary experience is met head-on with the continuity argument: the realness is not distinctive in the way the objection assumes; it is what the predictive loop does whenever it runs. In Section IX, a new empirical prediction is added: VR manipulations of multisensory consistency, interoceptive feedback, and top-down predictive weight should produce dissociations between presence and the metacognitive reality judgment that track RSP’s architectural decomposition. And in Section X, the entry on Barrett’s constructed emotion theory gains a paragraph on affective realism as an independent convergent line of support.

The paper does not solve the hard problem. It removes a category mistake about what the problem was. The brain has no switch for real. It has a recursive predictive loop, and the loop’s operation is what realness consists in — in waking life, in dreams, in VR, in DPDR, in the rubber-hand illusion, and in every other case where the architecture has been studied. The question of whether some experiences are really real, beyond what the loop is doing, was never a question about the brain. It was a question about a category the brain does not have. Singleton and Fotopoulou show it empirically. RSP says that is exactly what the constitutive identity thesis predicts, and that the dissolution of the hard problem follows from taking the showing seriously.