The headlessness argument I responded to earlier this week, in Ricky Williamson’s essay drawing on Douglas Harding, leaves a question I did not fully press in the reply. If the self-model has no architectural pathway to visualize the head from inside its own perspective — if the visual self-object is structurally absent from first-person experience — then what, exactly, is doing the work of feeling like a single continuous “me” from one moment to the next? You point at your face and find a void. You wake up tomorrow and find yourself. How do those two observations sit together?
The answer the RSP architecture suggests is unflattering to the everyday picture of identity. The self that feels continuous through time is not a single thing that persists. It is a series of architectural achievements stitched together by a narrative that runs in the present tense and reaches backward through memory. The selves being stitched are genuinely different selves — built at different ages, on different schemas, with different conceptual vocabularies, in different social contexts. The continuity is not preserved in the substrate. It is being constructed, continuously, by the part of the architecture that tells stories.
This is not a metaphysical claim about anatta or no-self in the strong sense. The self is real. The selves are real. What is constructed is the unity. And the unity is what we usually mean by “me.”
What Gets Installed First: The Named Self
Williamson’s observation that we are trained from birth to prioritize the third-person framing of ourselves is psychologically accurate, but the architectural reading sharpens it. The first thing that gets installed in the developing self-model is not the spatial first-person self. It is the social self — the self as named, addressed, described from outside.
An infant is given a name before it can use one. It is talked about before it can talk. It is photographed, described to relatives, predicted (“he’s a fussy sleeper,” “she’s a careful eater”), and corrected toward a stable identity that exists in other people’s minds before it exists in its own. By the time the child is able to articulate anything like an “I,” the I being articulated has already been thoroughly shaped by other people’s third-person model of who this child is. The child becomes a “you” before becoming an “I.”
In RSP’s architecture this is not a cultural accident. It is what the machinery is built for. The Level 3b recursive self-model uses the same neural circuitry that models other minds — the temporoparietal junction and superior temporal sulcus systems that compute what others are aware of, what they intend, what they expect. This circuitry is recruited and turned reflexively inward to model the self, which means the first thing it can model the self as is precisely the kind of object it can model other selves as. The self-as-social-object is the natural output. The self-as-first-person-perspective is a downstream achievement layered on top, not the starting point.
This is why headlessness feels surprising. The first-person spatial self — the headless body, the void where social convention placed a head — was installed onto a social self that had been running for years. The seam between the two is invisible in ordinary life because the integration mechanism papers over it. Harding’s pointing experiment makes the seam briefly visible. The void is what was there before the social self was installed; the social self has been overwriting it ever since.
What Gets Installed Last: The First-Person Spatial Self
The developmental sequence in Section V of Beyond the Bat — body schema present at birth, spatial schema by three to four months, affective-homeostatic regulation by six to nine months, attention schema by nine to twelve months, self-schema and mirror self-recognition by eighteen to twenty-four months, explicit theory of mind by four to five years — is not a smooth unfolding of a pre-existing self. It is a series of architectural commissionings in a strict order, each building on the last. The cascade-control proof in Appendix D requires that inner loops stabilize before outer loops can be added; the developmental ordering follows that constraint.
The child at eighteen months who first recognizes itself in a mirror is, in any architecturally meaningful sense, a different self from the child at three weeks. The child at four years who first reliably attributes false beliefs to others is a different self from the child at eighteen months. The adolescent who first feels their identity as a stable narrative across years is a different self from the four-year-old who passes the false-belief task. Each transition is an addition of architectural machinery that did not exist before. The previous self is not preserved; it is built on top of, partially overwritten, sometimes contradicted.
Adult self-experience inherits all of these layers. The first-person spatial self — what Harding pointed at and found absent — is the late, partial, structurally-incomplete layer installed onto a social-prediction substrate that does most of the actual self-modeling work. The headlessness is the experiential signature of which layer was installed when. The void is older than the social self. The social self is older than the narrative self. The narrative self runs on top of everything and gives the whole stack its felt continuity.
Many Selves Over Time
This means the “you” reading this paragraph is not the same self as the “you” who started reading this post. That sentence sounds extreme, and it is — the integration mechanism is doing its work right now, generating the felt continuity that makes the claim seem absurd. But pause and check: what specifically connects the two selves? The body, yes. Some memories from a minute ago, yes. A thread of attention through the paragraph, yes. But the actual neural state, the active beliefs, the precision weights on prediction errors, the contents of working memory — these are different from what they were sixty seconds ago. The architecture has updated. The previous state is not preserved.
This is true on every timescale. The you at six is not the you at twenty-six, even though the integration narrative says they are. There is no neural pattern, no architectural feature, no persistent memory representation that survives those twenty years intact. There is only the continuously-updated narrative that ties them together by reaching backward through reconstructed memory and forward through projected commitment. The continuity is the story. The selves are real; the story is real; the unity is a construction.
Context-dependent selves make the multiplicity more obvious. You-at-work and you-with-old-friends and you-alone-at-3am are not three masks worn by one underlying you. They are three different configurations of the Level 3a conceptual self-model running on partially-different cached schemas, with different priors active, different precision weights on different prediction errors. The friend who knew you in college and the colleague who hired you last year really are encountering different selves — the integration mechanism in your head insists they are encountering the same one, but the architectural fact is that they are not.
The Story That Integrates
The integration mechanism is Level 3a in RSP’s vocabulary — the conceptual categorization layer that uses linguistic and cultural categories to organize prediction errors at Levels 1 and 2 into coherent narrative chunks. Among other things Level 3a does, it categorizes selves-at-different-times as instances of the category “me.” This categorization is what feels like personal identity through time.
The categorization is constructive, not perceptual. It is not that there is a continuous self that Level 3a perceives and reports; there are discontinuous selves, and Level 3a categorizes them under a shared label, generates a backward-reaching narrative that explains how the earlier ones became the later ones, and runs that narrative as the foreground content of identity-experience. The narrative is updated continuously in the present tense. Tomorrow’s narrative will revise today’s in small ways and leave the revisions invisible. The you of tomorrow will sincerely report that they have always been the same person they are now.
This is not skepticism about identity. The narrative is a real cognitive achievement, with real causal powers, generating real continuity in behavior over years. Calling it constructed does not diminish it. What it does is locate it: the felt unity is architectural, not metaphysical. It is something the system does, not something it has.
The Seams Are Visible if You Look
The integration is normally invisible. Three recent reply posts on this blog touched on conditions in which it becomes visible.
Nick Chater’s choice-blindness experiments, drawing on the work of Johansson and Hall: subjects choose between two faces, are handed the wrong card via close-up magic, and fluently justify a preference for the face they did not choose. The integration mechanism is generating a coherent narrative without access to the actual cause. The seam between what the system did and what the system reports having done is exposed; the integration mechanism fills the gap before the subject notices it.
Gazzaniga’s split-brain interpreter, in the same Chater piece: when the corpus callosum is severed and the left hemisphere has no causal access to actions initiated by the right, the left hemisphere fabricates a confident, contextually appropriate, completely wrong explanation. The integration mechanism does not say “I have no idea why I did that.” It runs as if it knew. The seam between hemispheres is exposed by surgery; the integration mechanism papers over it anyway, with results that would be funny if they were not also evidence about how the integration normally works in intact brains.
Williamson’s headlessness, drawing on Harding: the pointing experiment exposes the seam between the social self and the first-person spatial self. The visual self-object that everyday self-experience assumes is present turns out to be structurally absent. The integration mechanism normally hides this; pointing at the void makes it visible for a moment, then social convention reinstalls the assumption that the head is there.
And the clinical dissociations in Appendix G — Capgras, depersonalization, alien hand syndrome, Cotard’s, dissociative identity disorder — are the integration mechanism failing in specific architectural ways. Each is a natural experiment that reveals the structure of the integration by breaking it at a specific joint. The Capgras patient’s “impostor” conviction is what happens when Level 1 affect mismatches Level 2 recognition and Level 3 cannot integrate the two. Depersonalization is what happens when Level 3b runs in spectator mode, decoupled from felt ownership. Cotard’s is what happens when the self-model generates the prediction “I do not exist” and cannot revise it. These are not weird edge cases. They are the integration mechanism revealing what it normally does by failing to do it.
Four independent pieces of evidence point at the same architectural fact. The self is constructed continuously, by a specific mechanism, out of materials (memory, prediction error, social-model output) that are not themselves selves. The integration is invisible when it works, visible when it fails. The continuity is the achievement; the multiplicity is what’s actually there.
What This Means in Practice
Chater’s practical move — self-authorship rather than self-discovery — gets its architectural justification here. There is nothing to discover because the integrated self is being constructed continuously. There is something to author because the construction is genuinely ongoing and partially under deliberate control. The Level 3a narrative mechanism is responsive to deliberate intervention: what you tell yourself, what stories you rehearse about who you are, what you commit to in advance — these feed back into the next round of integration.
This is also why some therapeutic interventions work and others do not. Cognitive therapies that target the narrative layer (Level 3a) can revise the integration mechanism; they work for problems that live at the level of how the self interprets and categorizes events. Somatic and trauma-focused therapies target Levels 1 and 2, where narrative revision cannot reach — the body schema and affective-homeostatic schema have to be re-integrated at their own level. The mapping from technique to level is interpretive rather than fully established by outcome research, but the architectural prediction is that a technique that targets the wrong level will fail for problems that live elsewhere. The same self-narrative revision will succeed for a Level-3 problem and fail for a Level-1 one.
The implication for identity-stability worries is also architectural rather than philosophical. The self does not need to be stable in order to be coherent. The narrative is the coherence, and the narrative is generated fresh in every moment. The you who feels different at fifty from how you felt at twenty is not betraying a stable underlying you; the underlying selves were always different, and the narrative is doing what it has always done. The work is not preserving a continuity that was always real. The work is authoring a continuity that has always been a construction.
Where This Leaves the Headless Body
Harding’s pointing experiment turned out to be doing more than one thing. It exposed the seam between the social self (installed first, doing most of the work) and the first-person spatial self (installed late and partially). It produced a moment in which the integration mechanism could not paper over the absence of a visual head. And it pointed, without saying so, at the larger architectural fact that the felt unity of self is always being constructed, just normally invisibly.
The previous reply post argued that the void Harding pointed at is the experiential signature of what the self-model architecturally excludes. This post extends that argument: the void is older than the social self, the social self is younger than we feel it is, the integration that holds them together is happening now. We were not born with a self. We were issued one in pieces, in a specific order, and the pieces are still being assembled. The headless body is not the truth behind the everyday self. The headless body is one moment in which the assembly is briefly visible.
We are not one self. We are many selves, integrated over time, building on the same story of who we think we are. The story is real. The selves are real. The unity is real in the way a building is real — it stands because of what holds it up, not because it was always standing.