There is a moment, in Douglas Harding’s work and in Ricky Williamson’s recent IAI essay drawing on it, where the reader is asked to do something simple. Point at objects in the room. Then point at your own face. From the first-person perspective — not from what you have been told is there, not from mirrors and photographs — what are you pointing at?
Harding’s answer is that you are pointing at a nothingness, a void, a world. There is no head there. The skull and face you have been assured exist do not appear in first-person experience. The world appears instead — on what Williamson, following Berkeley, describes as a two-dimensional screen where you were once told a head would be. To be a human being, on this reading, is to be a headless body with a world coming out of your shoulders.
I think this is correct as a phenomenological observation. I think it is also evidence about the structure of the self-model that RSP did not seek but should have predicted. The conclusion Williamson draws — that the self is therefore an illusion, that subject and object are metaphysically one, that you do not have a head — is a stronger reading than the data require. The architecture explains the void without making the head disappear.
What Williamson and Harding Get Right
The agreement is substantial and the agreements matter. Three points are load-bearing.
The headlessness experiment is a real phenomenological datum
The pointing experiment is not a trick or a verbal puzzle. When you point at your own face from inside your own perspective, what you find is not a head. You find the finger, the room behind the finger, perhaps two faint nose-shadows, and a world that opens out where common sense told you a skull would be. Williamson is right that this is surprising the first time and remains surprising on repetition. Harding’s point is genuine, not metaphorical: the first-person experience of pointing at your own face is not the experience of pointing at a head-shaped object.
Williamson is also right that this is hard to see because we have been trained from birth to prioritize the third-person framing of ourselves. Babies are given names, turned into objects, addressed as something separate from the world they experience. The pointing experiment is unusual because it returns first-person evidence to the privileged position it occupies before social convention overwrites it.
The self is not a substance
Williamson endorses what Wittgenstein puts cleanly in Tractatus 5.631: “The thinking, presenting subject; there is no such thing.” The self, in the substantial Cartesian-theater sense, is not findable. Look for the homunculus behind the experience and you do not encounter it. RSP makes the same anti-substantialist move from the opposite direction: the recursive self-modeling loop at Level 3b is a dynamic process, not a hub or a viewer or an inner agent. The loop has no fixed center; the center is the loop running.
This shared rejection of self-as-substance is the deepest point of agreement, and it is rare. Most contemporary theories of consciousness either ignore the question or smuggle in a residual subject somewhere. Williamson and RSP both refuse the substantial subject from the start.
Mach’s self-portrait is the right kind of evidence
Ernst Mach’s 19th-century self-portrait — drawn from inside his own visual field, showing the body trailing off below, the nose and brow as faint edges, and the room beyond — is one of the cleanest pieces of first-person phenomenology ever produced in figural form. Williamson is right to foreground it. The image is a snapshot of what the visual self-model excludes: the head as a visual object. Mach noticed it. Harding noticed it. Williamson reminds us that we mostly do not notice it.
Both Williamson and RSP reject the self-as-substance. Both treat first-person evidence as primary data rather than as testimony to be discounted. Both find the everyday social-convention third-person framing of the self — the named, photographed, described-from-outside self — to be obscuring something real about first-person structure. The disagreement is not about whether headlessness is a genuine first-person observation. It is about what the observation entails.
Headlessness Is What the Architecture Predicts
Here is the part Williamson and Harding do not say, but that RSP supplies. The self-model at Level 3b is built from social-prediction machinery — the same theory-of-mind circuitry that models other minds, recruited and turned reflexively inward. The body schema at Level 2 provides proprioceptive self-representation: where the limbs are, where weight presses, what the body is doing. Neither of these systems generates a representation of the self as a visual object seen from inside its own visual field. There is no architectural pathway that would put a visual head into first-person experience. The void Harding points at is the experiential signature of this structural absence.
Vision is anchored at the eyes. The visual field extends from the eyes outward into the world, with the body trailing off below as a fading proprioceptive-visual hybrid. The Level 3b self-model represents the self as others would see it — the social self, the named self, the responsible-agent self — but it does not include a visual rendering of the head from its own observational origin. The head is precisely what cannot appear in the field, by the geometry of the architecture.
This means Harding’s pointing experiment is not just a curiosity. It is an architectural probe. It selectively makes salient the gap between the social-object self-model (Level 3b) and the perspectival origin of experience (anchored at the eyes, with the head as a structural absence). The experiment functions much like the clinical dissociation cases catalogued in Appendix G of Beyond the Bat — Capgras, depersonalization, alien hand syndrome — except that it is non-clinical and any subject can perform it. It reveals the topology of the self-model by exposing what the self-model does not contain.
This is a candidate empirical anchor RSP does not currently exploit. Appendix G’s twenty-eight dissociation entries catalog what goes wrong when the architecture breaks. Harding’s pointing experiment shows what the architecture excludes when the architecture is working normally. Both are evidence about the structure, but the second has the advantage that it does not require pathology.
Where Williamson and RSP Genuinely Disagree
The disagreement runs through a single sentence in Williamson’s essay:
“the first-person perspective is the only real perspective that anyone has ever known, and so we must take this to be the only truly real one and of primary importance when we are looking for answers.”
This is the strong methodological claim: the first-person is ontologically privileged; the third-person is social convention. Williamson is honest enough to acknowledge the incommensurability this creates. He admits that he must take third-person evidence on board — he does have a head, mirrors and other observers confirm it — but he insists that from the only real perspective, the first-person, the head is not there. Two descriptions of one body that cannot be reconciled.
RSP makes a different move. First-person and third-person are not ranked perspectives. They are perspectival pairs that co-refer to the same underlying architecture. The first-person view is what the recursive self-modeling process is like from inside the loop. The third-person view is the same process described from outside the loop. Neither is more real; neither overrides the other. Both are descriptions of one event in two vocabularies that necessarily co-exist because the system is the kind of system that admits both descriptions at once.
This dissolves the incommensurability Williamson acknowledges but cannot resolve. The head visible in the mirror and the headlessness visible in first-person experience are not in competition. They are different aspects of the same body described from different architectural standpoints. The mirror shows the body as a social object (which the third-person self-model represents). The pointing experiment shows the body as a self-object (which the first-person self-model represents). Both are accurate within their domain. Neither is the truth that the other is missing.
No-Self vs No-Substance
Williamson endorses the strong reading: the self is an illusion. The headlessness experiment reveals that there is no subject, and this connects (via Wittgenstein and via the Sam Harris lineage Williamson invokes at the end) to the Buddhist non-self tradition. Self-as-illusion is the destination.
RSP gets off the train one stop earlier. The self that is “empty” in the substantial, fixed, independent sense is correctly diagnosed as illusion. But the self as a dynamic process — the recursive self-modeling loop running at Level 3b, with the cascade-control structure described in Appendix D — is real and functional. The loop has no hub, no fixed center, no homunculus inside. But the loop is the self, and the loop has genuine causal powers, predicts and adjusts lower-level processing, generates the phenomenal life of the creature whose loop it is. The absence of the self-as-perceived-object does not entail the absence of the self-as-process. Headlessness reveals that the self-schema does not represent itself as a visual object, which is true and architecturally important, but it does not reveal that no self-process exists. The Level 3b loop is causally real even when no self-object is perceived.
The distinction matters for what one does next. If the self is an illusion in Williamson’s strong sense, the practical upshot is roughly contemplative — recognize the illusion, dwell in the headlessness, dissolve. If the self is a process in RSP’s sense, the practical upshot is architectural — understand which kinds of self-modeling are working, which are failing, and what specific clinical conditions reveal about the structure. Both are legitimate orientations. They are not in contradiction so much as in different relationships to the same underlying observation.
Subject-Object Non-Duality, Without the Berkeley Move
Williamson goes further: “I am my world.” The Wittgenstein Tractatus 5.63 quote is invoked. The subject and the object are united at the shoulders, conjoined in a single perceivable unity. Berkeley’s point that distance is never directly perceived is recruited to argue that the world collapses onto the plane where the head would be, with no spatial separation between self and world.
RSP gets to a closely related non-dualism by a different path, and arrives at a weaker version. Subject and object, on the perspectival-pair reading, are not metaphysically merged. They are two descriptions of one architecture. The subject is the architecture from inside; the object is the architecture from outside. They are not separated — the gap Williamson rightly diagnoses as illusory was never real — but they are not merged either. They are not the same kind of entity; they are different stances toward the same set of events.
RSP does not need the Berkeley/2D-screen argument to dissolve the subject-object gap. The architecture handles it directly. Spatial experience is a predictive model the system builds at Level 2, not a direct readout of metric geometry. Distance is constructed, as Berkeley noticed, but the constructedness does not entail that there is no distance — only that the experience of distance is a model output rather than a perception of an independent fact. The phenomenological observation Williamson grounds his stronger metaphysical claim on is real; the stronger metaphysical claim is not entailed by it.
| Williamson (headlessness) | RSP (recursive self-modeling) | |
|---|---|---|
| The self is | An illusion; not findable in first-person experience | A real dynamic process at Level 3b; not an object, but a loop |
| Headlessness shows | We have no head; subject and world are one | The self-model does not represent the head as a visual object — architectural fact, not metaphysical claim |
| First-person status | The only truly real perspective; third-person is social convention | One of two co-referring perspectives; neither is more real than the other |
| Subject and object | United at the shoulders — metaphysical non-duality | Two descriptions of one architecture — perspectival pair |
| Free will | Illusion; follows from non-duality | Architecturally graded agency at Level 3b; no commitment to libertarian free will |
Free Will, Briefly
Williamson’s essay raises the free-will question at the end and gestures at the conclusion that free will is illusory — if there is no gap between self and world, the asymmetry that would ground special control over the body collapses. RSP’s Section VII on philosophical implications does not engage free will directly, but the cascade-control structure described in Appendix D is not silent on agency. Level 3b self-prediction genuinely modulates Level 2 and Level 1 processing; the system represents itself as having predictive influence over lower levels and the representation is accurate, because the influence is causally real.
This is closer to compatibilism than to Williamson’s harder conclusion. There is graded agency at Level 3b without any commitment to libertarian free will. Whether this is enough to count as “real” free will is a question the paper does not answer, and Williamson’s essay does not engage in enough detail to make the disagreement sharp. Genuine open question rather than resolved contradiction.
The Pointing Experiment as Architectural Evidence
The most useful thing RSP can take from Williamson is the pointing experiment itself. Appendix G of Beyond the Bat catalogues twenty-eight clinical dissociation entries across eight tables — cases where specific architectural breakdowns reveal the seams of the self-model. Capgras shows what happens when Level 1 affect mismatches Level 2 face recognition. Depersonalization shows what happens when Level 3b runs in “spectator mode” decoupled from felt ownership. Split-brain confabulation shows what happens when Level 3b has structurally no causal access to its own actions.
Harding’s pointing experiment is none of these. It is not pathological. It does not require neurological damage or special training. Anyone can perform it in a few seconds. What it reveals is what the working architecture normally excludes — the visual self-object that the self-model does not generate, the head-as-seen-from-inside that has no architectural pathway to first-person experience. This is a non-clinical architectural probe, and RSP’s Section IX (Empirical Predictions) currently has none of these. Adding the pointing experiment — or a class of related phenomenological probes — as a category of evidence alongside the clinical dissociations would close a methodological gap. The architecture predicts that no first-person visual head appears; the prediction is verifiable in seconds by any reader; the verification is a direct empirical anchor for Level 2 body-schema and Level 3b self-model structure.
Among the frameworks compared in Section X of Beyond the Bat, Williamson and Harding sit most naturally in Family H (Phenomenological Theories) — alongside Gurwitsch’s theme-field-margin structure already engaged there. The framework shares with phenomenological traditions the commitment to first-person evidence as primary, the rejection of third-person reductionism, and the dissolution of the hard problem through perspectival analysis rather than mechanistic explanation. Its non-dual resonances connect it to contemplative frameworks (the Sam Harris lineage Williamson notes), but it is not reducible to them — Harding’s method is explicitly empirical and first-person experimental rather than doctrinal. The recommended classification is complements: Williamson supplies a first-person phenomenological datum — the structural absence of the self-model’s own observational origin — that RSP’s architecture can accommodate and explain. RSP supplies the mechanistic and developmental scaffolding that Williamson’s account defers. Beyond the Bat should cite Williamson 2026 (IAI) alongside Harding 1952 (The Hierarchy of Heaven and Hell) in the Family H subsection, with a note that headlessness is the experiential correlate of what RSP formalizes as the topology of the self-model at Level 3b.
Williamson is right that you have no head when you point at your face from inside your own perspective. He is right that the self is not an object you can find. He is right that the social-convention third-person framing has trained us out of noticing how strange the first-person actually is. What he leaves untheorized is why the architecture would produce these phenomena in just this form. The void is real. The void is also structured. The structure has timescales, levels, and dynamics that can be measured, perturbed, and predicted. Headlessness is not the end of the story. It is the first piece of architectural evidence visible to anyone who knows where to point.