Stripping The Self

Who am I? Who is anyone? Am I thinking, or observing thinking? Many questions like this plague many minds, and even the very fact of this commonality — that we are all drawn to them — is itself suspicious to me, and interesting. I wonder what the self would be if we tried to eliminate all the pre-conceptions and assumptions that have built up around it.

So, what am I? Am I my house, my car, my watch, my money? Does losing any of those mean the self is eliminated? Surely not. We can imagine that no possession could be attributed to our self, since we can eliminate them and still survive.

Then, what am I? Am I a salaryman? A director? A janitor? A parent if I have children? A loser in some views and a winner in some views? I doubt it. Someone fired in a layoff is still themselves — albeit perturbed. Someone can have a long career and drop it and still be themselves. Any such role we can imagine can be eliminated, and the self survives.

What’s left to define me? My nationality, which is given? My name? Many others shared, share, and will share it. My political affiliation? My ID card? No matter how much we iterate, no type of external identification can be attributed to the self, because the self survives them all being eliminated.

No possessions, roles, or external views survive being eliminated. What’s left is to turn inwards. When we look inwards, what is the most bearing thing we observe? Our own memories. Would we still be a self if we removed all our memories?

The literature points elsewhere. Memory loss damages the self without eliminating it.

Several cases illustrate this. Korsakoff syndrome — severe anterograde amnesia, where patients are unable to form new memories yet still retain their personality, preferences, humor, and still experience life. General anesthesia, where a person wakes up with hours of gaps in their memories yet their self persists. Deep dreamless sleep, much more common: just like general anesthesia, big gaps in memory, yet we wake up every day as ourselves. And more severe cases like dementia and Alzheimer’s, where caregivers and family keep reporting that “someone” is still there, sometimes showing up in very small briefs of time. Memory is constitutive of parts of the self — the narrative thread, autobiographical continuity, recognition of loved ones — but not the self itself. The self can be perturbed by memory loss without being eliminated by it.

But then, what else can we eliminate? There are still some ideas we can identify with. Am I me because I am a warm or cold person? Or an open or closed mind? If I am a conscientious human or not? Am I defined by the food I like, the music I enjoy, the partners I choose, or the lifestyle I live? Doubtful. If not even that, what else? By the God I choose to believe in? By the ethics and morals that I obey? By my own self-beliefs like “I’m a good person”? Am I my habits or my emotional defaults?

Many cases show a self still persists across all variations of everything I just mentioned. From religion we know of the conversion of Saul of Tarsus (~33–36 CE), who persecuted Christians but, upon a certain event, became an apostle and wrote much of the New Testament. Same body, same “self” supposedly, but fully opposed values. We know of Phineas Gage (1848), a railroad foreman from the United States whose frontal lobe was pierced by a large iron rod in an accident; he survived 12 more years, with a sudden and dramatic personality shift that his family and coworkers described as “no longer Gage.” After many years he managed to recover substantially and reclaim much of his prior personality. The self was damaged, and over time it was healing — if “healing” is the right word at all. Finally, take a long look inside yourself and ask: how do I view myself now compared to when I was 18? Or 10? Or how I would be at 50, or 80? Am I the same across all time? Surely not. Each period has its own context which induces and influences very many of our choices and beliefs and personality and “self.” What drives the context? The state of the whole world and reality, it might not be an exaggeration to say. Under this view we can also ask: what HAS remained invariant across all times and my “self” at 10, 18, 30, 50, 80? Eliminate all that changed and varied across your whole life, and see what was and is the only invariant present. So we observe that personality traits, preferences, beliefs of any kind, and all habits of thought and emotional defaults can vary. All of these change. Some change slowly, some flip overnight. But what I am still walks through them all — in my view, they can be eliminated and there is still a self left, whatever it may be.

But then, what else can we eliminate? What else do we have? A body. Can we eliminate the body? Not strictly speaking, but here we can use a classical conundrum to help us think: the ship of Theseus. The main idea: the ship Theseus (hero in Greek mythology) used in his adventures was later preserved by the Athenians; over time every plank rotted and was replaced. By the end, no original part remained. The question: is it the same ship? This subject has been dissected many times. Hobbes’s extension (1655): now imagine all the discarded planks are reassembled into a ship. Two ships exist. Which is Theseus’s?

In the same way we can ask of our body: at what rate are our “planks” being replaced? We know the cells in our body change, but how much exactly, and how? In 2005 the rates were quantified directly via retrospective birth-dating of human cells using atmospheric C-14 from above-ground nuclear tests (Spalding, Bhardwaj, Buchholz, Druid, Frisén, Cell 2005) — intestinal lining ~5 days, skin ~2–4 weeks, red blood cells ~120 days, most cells in months to years. A major exception: cortical neurons, which mostly persist through the whole life. The net effect — across roughly 7–10 years, most of the atoms making up our bodies are replaced. Biology instantiates Theseus’s ship at the cellular and atomic level. The person persists across complete material substitution. The cortical neurons matter particularly here, since they largely persist for life and form the synaptic patterns that store memory, personality, and body schema. So even if 99.9% of the body is replaced, one might argue that the 0.1% that isn’t is the self. This is true, but it just means we need to move the elimination operation inside the neurons themselves — cells may persist as organizational units, but the matter inside them doesn’t. Proteins are constantly turned over, the membrane is continually adapting, even the mitochondria are replaced. The cell persists; its interior doesn’t. What happens within the cortical neurons themselves is the same dynamic we observe in the rest of the body, just inside another boundary.

Let’s turn to something else that helps here: amputees. A large portion of them experience phantom sensations from the missing limb — documented in Ronald Melzack, “Phantom limbs and the concept of a neuromatrix” (Trends in Neurosciences, 1990). The brain’s body schema persists even across removal. A particularly well-known experiment, the mirror box for visual feedback, was devised by V.S. Ramachandran (developed with Rogers-Ramachandran 1996; described accessibly in Phantoms in the Brain, 1998), and demonstrated that the way the self models the body is plastic and can be manipulated externally. Ramachandran & Hirstein, “The perception of phantom limbs” (Brain, 1998) showed that after amputation the somatosensory cortex reorganizes and the area that formerly handled the missing limb gets recruited by adjacent areas. The self adapts, the body model rewrites itself, the person persists.

Consider locked-in syndrome — when a patient is fully conscious but has lost almost all voluntary motor control, usually caused by a brainstem stroke, traumatic brain injury, or late-stage motor neuron disease. The term was first coined by Plum and Posner in The Diagnosis of Stupor and Coma (1966). It is different from coma, persistent vegetative state, and minimally conscious states — locked-in patients are fully aware and cognitively intact. The body becomes completely restricted and goes offline; the mind does not. A close cousin is Stephen Hawking, diagnosed with ALS at 21 with a prognosis of two years. He reached 76 before dying, contributing world-class theoretical physics in between. ALS is not strictly locked-in syndrome, but late-stage ALS approaches it: by his last years Hawking communicated only via a single cheek-muscle twitch. The body was effectively a life-support system for a still-active intellect for many years.

All these examples show that significant chunks of the body can go without eliminating the self. We must be careful not to claim the body doesn’t matter — none of the prior examples, nor any other we can conjecture, show that the body as a whole is irrelevant. Even Hawking still had a brain, a heart, blood pumping. A boundary as a whole is a necessity, not a quirk.

Having stripped personality and body parts, let’s ask what remains of the self that we can still eliminate. Am I the same person when I am drunk? When I am drugged? When I am overcome with rage? When I am overcome with peace? Am I still me if I no longer eat the same? Am I still the same me if I started doing something new? Each one has an answer. Let’s relate to what we know exists. We know that sleepwalking, sleep talking, and REM behavior disorder exist. People walk, eat, drive, attempt sex, and occasionally commit violence while asleep. The body is definitely “you,” and the actions performed fail to be registered by the conscious self. The famous case from Toronto: in 1987 Kenneth Parks drove roughly 20 km while asleep, killed his mother-in-law, severely injured his father-in-law, and drove himself to a police station. He was acquitted of murder on automatism grounds. Whatever drove the body while the conscious self was not there to be aware is, most definitely, not the self.

What about drugs? All of them alter, in different ways, whoever we are. Alcohol, anesthesia, psychedelics, sedatives, stimulants. Behavior may change radically; values can invert briefly. There is a common phrase used everywhere: “you weren’t yourself last night.” A cultural admission, across time, that the same body produced behavior the self negates.

What about self-altering mind phenomena — depersonalization (you experience yourself as not-yourself), derealization (where the world stops appearing real), dissociative fugue (extended unaccountable behavior with no memory)? Different alterations produce dramatically different behavior in the same body.

Finally, the classical example of rage. Once called “heat of passion,” a doctrine in Anglo-American law that has for centuries recognized that behavior in extreme emotional states is a different moral category — it can mitigate murder to manslaughter. Depending on the emotional state we inhabit, the brain produces self-incompatible decision-making.

Across all of these, the same distinction is engraved: “I wasn’t myself.” This only makes sense if there’s a normal “you” distinct from the actions of your body in altered states.

Where else should we look to eliminate? We can go deeper than where we went when we discussed memory and concluded it damages but doesn’t eliminate the self. There is an implicit assumption there: that during the memory-loss period the self existed at all. Did the self exist during anesthesia? During dreamless sleep? Was there really anything there present?

We previously used anesthesia as evidence that memory can vanish without the self vanishing. The question now is different: was a self present during the period the patient was anesthetized, or did the experiencer simply not exist? Most contemporary anesthesiology research suggests the latter — under deep general anesthesia the typical neural signatures associated with consciousness (gamma synchrony, frontoparietal integration, cortical effective connectivity) disappear. Massimini et al., “Breakdown of Cortical Effective Connectivity During Sleep” (Science, 2005), used TMS-EEG to measure the complexity of the brain’s response to a magnetic pulse during wakeful consciousness and during non-conscious states (deep NREM sleep, anesthesia, coma). The result is telling: the first has high complexity, the second collapses to low levels. And — not necessarily for anesthesia only, but for our own dreams as well — we wake up disoriented, which is itself suggestive. If the thread were continuous, the resumption should feel seamless and frictionless. The fact that it requires any effort at all to re-assemble what “now” is hints that the thread is being rebuilt, not extended.

Let’s take dreams more in-depth. We do this every night. Most of us get around 1–2 hours per night of slow-wave deep sleep with no experiential content. You wake up in the morning and have no idea exactly when you fell asleep — there’s basically just a blank darkness from then until now. Plus REM gaps without dream recall. Across a single life of roughly 25,000 nights, that’s tens of thousands of hours during which no experiencing being is present. Yet we wake up every morning convinced of continuous selfhood. The Massimini result generalizes. NREM stages 3–4 show perturbational complexity indistinguishable from anesthesia or coma. Whatever consciousness is, by the integrated information measure, it isn’t running during deep sleep. And insomniacs frequently report having been awake all night when EEG shows extended sleep. The phenomenon goes the other way too — people misperceive how long they’ve slept. This divergence between measurable time-asleep and felt time-asleep is hard to reconcile with a continuous experiential thread; the felt amount of time is being constructed from whatever information is available, not retrieved instantly from a shelf.

So we’ve established that the continuous thread isn’t preserved across gaps, but rather generated each morning by the brain retrieving memories and assembling the sense “I am the same as yesterday.” The thread is stitched at each end, not threaded through. Every time you recall a memory, you reconstruct it (Nader, Schafe & LeDoux, Nature 2000; Lee, Nader & Schiller, Trends in Cognitive Sciences 2017), and the neural representation is altered during recall. So even the parts of memories that “prove I’m the same person every day” are being changed continuously. Time perception itself is malleable — subjective time slows in emergencies, expands under psychedelics, compresses in flow states (Eagleman, The Brain: The Story of You, 2015). A continuous thread shouldn’t be that elastic. One last striking example is Michael Gazzaniga’s work on patients with intractable epilepsy whose corpus callosum was severed. When the right hemisphere produced an action and the left hemisphere was asked to explain it, the left fluently invented plausible-but-false reasons. Gazzaniga called this the “interpreter module.” The brain is constantly generating coherent narratives post-hoc, even when the underlying facts have no coherent source. The dynamics of stitching the narrative together is always running in everyone, not just split-brain patients. This contradicts how we feel inside. From inside, time feels continuous. But the felt continuity is being constructed at each waking moment from the memories and impressions the brain currently has access to. There’s no separate “thread” of you that is present across the gaps; there’s a brain that, each morning, generates the sense “I am the same as yesterday” by retrieval and reconstruction. Even the continuous experience of being myself is a construct happening dynamically.

So memories and the felt continuity of who we are can both be eliminated, and still something like a self persists. Where can we move next? Can there be a self without a brain, or even a body at all? Surely not. Or is it so surely? Let’s reach for some classical thought experiments.

The brain in a vat. Hilary Putnam, Reason, Truth and History (1981). The setup: imagine you’re a brain in a vat, fed sensory input by a sufficiently capable computer; you wouldn’t be able to tell. Putnam actually used BIV for an argument about meaning and reference (his anti-skeptical “we can’t be brains in vats” move), not for embodiment per se. The implication for us: if the brain alone (no body) could preserve the self, then specific biological embodiment isn’t constitutive — the body was a way to provide and maintain sensory input, not the self itself. We’ve already argued that the self can survive partial body elimination; this is the extreme limit of that argument. Some would say the brain alone preserves the self; some would say the missing body would make it a different “me” than they think they are.

Whole brain emulation pushes the same question further. The mainstream functionalist position: if a sufficiently detailed simulation of a brain runs the same computations as the original, it instantiates the same self. The literature is dense — Hans Moravec, Mind Children (1988); Ray Kurzweil, The Singularity Is Near (2005); Robin Hanson, The Age of Em (2016), where Hanson works through the economic implications of em-society in detail. The most rigorous technical treatment is Sandberg & Bostrom, “Whole Brain Emulation: A Roadmap” (Future of Humanity Institute, 2008).

The opposing positions have not gone away. Searle’s Chinese Room (Behavioral and Brain Sciences, 1980): syntactic manipulation isn’t semantic understanding, therefore a simulation isn’t conscious. Heavily contested, but the intuition behind it doesn’t dissolve under counter-arguments — the argument keeps re-appearing because something in it isn’t easily eliminated. 4E cognition (Andy Clark, Shaun Gallagher, Evan Thompson, descended from Maturana and Varela) holds that cognition isn’t in-the-head; it’s distributed across brain, body, environment, and tools. Uploading the brain wouldn’t preserve the cognitive system, just one component of it. Penrose-Hameroff goes further, arguing that consciousness involves quantum effects in microtubules that classical computation can’t simulate. Highly speculative and fringe in mainstream neuroscience, but it represents a serious anti-upload stand from yet another direction.

Empirically, where are we? Far away. C. elegans (~300 neurons) has been fully mapped since 1986; OpenWorm has been simulating it for years with ambiguous results. The female Drosophila brain connectome was published in 2024 (FlyWire collaboration, Princeton/Cambridge/HHMI). The MICrONS project published the largest mammalian dataset to date in Nature 2025 — about 1.4 mm³ of mouse visual cortex, with on the order of 200,000 neurons reconstructed and ~500 million synapses mapped. Humans are many orders of magnitude beyond that, probably decades away. Why? Because the empirical question of whether the wiring diagram is sufficient for behavior — let alone for the “self” — remains open. We don’t know what’s essential.

Then there is teleportation. Derek Parfit, Reasons and Persons (1984), Part III. The setup: a teleporter scans your body, destroys it, rebuilds an identical copy at the destination. Is the rebuilt person you? Now imagine the original isn’t destroyed — there are two of you. Which is the original? Parfit’s argument: identity is not what matters. What matters is psychological continuity, and continuity can branch, gradate, and split. The case is poignant either way you take it. If you accept that the teleporter preserves you, you’ve conceded that embodiment in the original body isn’t the self. If you reject it, you have to specify what’s missing, and Parfit’s point is that no specification works.

Close to the teleporter is gradual replacement, the Theseus argument run on neurons. Replace one of your neurons with a functional silicon equivalent. Are you still you? Almost everyone says yes. Now another. And another. Continue until every neuron has been replaced. At what point did you stop being you? Either no point — gradual replacement preserves the self all the way through, in which case upload preserves it — or some specific point, which would be arbitrary. Either answer pushes you somewhere uncomfortable. Moravec used the argument in Mind Children (1988); David Chalmers gave the careful philosophical treatment in The Conscious Mind (1996), particularly the “fading qualia” and “dancing qualia” thought experiments.

Practical situations have come out of these thought experiments too. Cryonics — Alcor Life Extension Foundation and the Cryonics Institute preserve heads or whole bodies in liquid nitrogen, betting that future technology can revive or upload them. Around 500 patients are preserved between the two organizations as of the mid-2020s. Scientifically uncertain, ethically contested. It is a real practice founded on the bet that the self is substrate-portable. Vitrification (rather than freezing) is a technical advance that makes the bet less obviously hopeless: cells preserved this way have been revived in some experimental settings, and Greg Fahy’s group at 21st Century Medicine demonstrated rabbit kidney vitrification with subsequent transplant function (Fahy et al., Organogenesis, 2009).

These thought experiments don’t resolve. The body can be modified, simulated, gradually replaced, or — in thought experiment — eliminated entirely, and our intuitions about whether the self survived are undecided. That undecidedness is itself the evidence. If the body were the self, we wouldn’t be able to imagine the self without it.

Where does this naturally lead us?

But self-modeling is a sophisticated cognitive achievement that arrived late in evolution and arrives late in human development. Most life on Earth doesn’t have it. Yet most life on Earth clearly has selves in the operational sense — systems that distinguish themselves from their environment, maintain their own boundary, act on their own behalf. So self-modeling can’t be what selfhood IS. There are selves without self-models — and they vastly outnumber selves with self-models on this planet. What survives the elimination of self-modeling is the system that does the distinguishing, even where there’s no representation of itself doing it.

Consider mirror self-recognition (MSR), the canonical test for self-modeling. Its origin is Gallup’s 1970 chimpanzee study (Science, “Chimpanzees: Self-Recognition”): anesthetize a chimp, mark its forehead with red dye, present it with a mirror. Mirror-naive chimps treat the reflection as another animal; experienced chimps reach up and touch the mark on themselves. The “mark test” became the gold standard. The current passing list is small — chimpanzees, bonobos, orangutans, gorillas, dolphins, Asian elephants, magpies, with manta rays in some preliminary work. The non-passing list is much longer: most primates and monkeys, dogs, cats, most birds, most fish, almost all invertebrates. All of these clearly have selves in the relevant behavioral sense — they protect themselves, eat, mate, fight, navigate, suffer. They lack the cognitive achievement of modeling themselves. So the cognitive achievement and the operational selfhood are different things. The model is one consequence of having a sufficiently complex brain; not a requirement for being a self. There are caveats, of course — failing the mirror test doesn’t necessarily mean lacking a self-model. Some species don’t engage with mirrors at all; others may have non-visual self-recognition (Bekoff 2001, and others, argue dogs may have olfactory self-recognition). The point doesn’t depend on the exact list. It depends on the asymmetry: selves without self-models are abundant, selves with self-models are rare, and operational selfhood predates and outlives the cognitive overlay.

Going further down the complexity ladder makes the asymmetry sharper. Bacteria use CRISPR-Cas systems to recognize invading viral DNA and remove it while leaving their own DNA intact (Doudna & Charpentier, Nobel Prize 2020). The system encodes “self” molecularly — sequences matching the bacterial genome are spared, sequences that don’t are destroyed. This is self/not-self distinction at the level of nucleic acid pattern, with zero cognitive content. The same bacteria, via quorum sensing (Bonnie Bassler’s work, summarized in “Small Talk”, Cell 2002), detect local conspecific density and switch behavior at thresholds — they have a “social self” without any cognitive representation of one. The system is self-relevant; the self-model is absent. The earliest articulation of this point is Maturana and Varela’s autopoiesis (1972 in Spanish; Autopoiesis and Cognition, 1980): a living system is a network of components that continuously produces the network of components that produces it. The cell maintains its own boundary against entropic dissolution. No mirror, no model — just the operation.

Plants run the same pattern. Gruntman & Novoplansky (PNAS, 2004) showed plants modify root architecture based on whether neighboring roots belong to the same plant or another plant — they distinguish self from not-self at the root level without any representation of selfhood. The distinction is operational and biochemical. Self-incompatibility in flowering plants (Igic & Kohn, PNAS 2001) is the same idea at the cellular level: pollen carries genetic markers, the stigma rejects matching markers, self-fertilization is prevented by molecular recognition. The plant immune system itself (Jones & Dangl, Nature, 2006) is self/not-self distinction with no nervous system, no representation, two layers of pattern-recognition.

Insects and simpler animals show the asymmetry from yet another angle. Honeybees navigate by sun-compass, communicate by dance, and bumblebees show striking cognitive flexibility — improving on observed solutions to a novel ball-rolling task (Loukola et al., Science 2017) — and yet, by the conventional view, don’t pass mirror tests and don’t model themselves explicitly. Their behavior is densely self-relevant: defending the hive, finding food for it, returning to it. Slime molds, single-celled organisms, solve mazes, optimize networks (Tero et al., Science 2010, the famous Tokyo subway map paper), learn from experience (Reid et al. 2016). No nervous system, no self-model, yet self-preserving behavior, learning, decision-making. The slime mold is the cleanest demonstration: self-relevant action without self-modeling is real, pervasive, and ancient.

Humans don’t escape the asymmetry; they show it developmentally. Mirror self-recognition emerges in human infants around 18–24 months (Lewis & Brooks-Gunn 1979). Before that age, the infant clearly has a self in any operational sense — protects the body, prefers the mother, has distinct reactions and preferences — but doesn’t yet model itself as a distinct entity in the mirror. Theory of mind emerges around four years (Wimmer & Perner 1983; Baron-Cohen, Leslie & Frith 1985 on the false-belief task), adding another modeling layer. The pre-four-year-old child is clearly a self; they just don’t model selfhood explicitly yet. And the model can detach in the other direction too — depersonalization disorder, certain dementias, severe anesthesia recovery all leave a system that protects itself, responds to its environment, and has continuity with its earlier self, but with the self-modeling part offline or distorted. The model can come and go without the operational self vanishing.

The most explicit contemporary statements of this distinction come from Thomas Metzinger and Anil Seth. Metzinger, in Being No One (2003) and The Ego Tunnel (2009), argues that the brain generates a phenomenal self-model and that no one is “behind” the model — the model is what selfhood, in the cognitive sense, IS. But the model is a model OF something, and that something is what we are after here. Seth’s “controlled hallucination” view (Being You, 2021) sharpens the distinction further: he separates the experience of being a self (the model) from the actual condition of being a system that distinguishes itself from its environment (the operation). The operation is the prior, the deeper layer; the model is what the operation produces in brains that became complex enough to model themselves.

So the self survives the elimination of self-modeling. The model is downstream of selfhood, not constitutive of it. Bacteria do the distinguishing molecularly. Plants do it biochemically. Insects do it behaviorally. Mammals add a model on top, primates extend the model with social inference, humans extend it further with autobiographical narrative. The model is layer upon layer of refinement. The operation underneath is the same one bacteria do.

Where else can we go? We’ve moved through concepts, biology, and philosophy. Let’s move now to first-hand evidence rather than thought experiments — to what people experience when the felt sense of being-someone-here breaks down. Three families of cases: psychedelic ego dissolution, meditative cessation, and pathological dissociation. They are produced by very different causes but they all do the same thing: they take the experiencer offline while something keeps reporting back.

Start with psychedelics. Ego dissolution under high doses of psilocybin, LSD, ayahuasca, and especially 5-MeO-DMT has been a credible neuroscience subject for roughly a decade now, mostly thanks to Robin Carhart-Harris’s group at Imperial College London (now UCSF). The work is interesting because it doesn’t just report what subjects say; it ties what they say to specific neural mechanisms. Carhart-Harris’s “entropic brain” hypothesis (Frontiers in Human Neuroscience, 2014) frames psychedelics as relaxers of the brain’s hierarchical priors — particularly the Default Mode Network, the cluster of frontal and parietal regions identified by Marcus Raichle’s group in 2001 as the substrate of self-referential thinking. The 2019 REBUS framework (Carhart-Harris & Friston, Pharmacological Reviews) goes further: ego dissolution is what happens when the high-level priors that construct the phenomenal self-model are temporarily released. Imaging studies confirm it directly — psilocybin reduces blood flow to medial prefrontal and posterior cingulate cortex (Carhart-Harris et al., PNAS 2012); LSD-induced ego dissolution scales with global functional connectivity (Tagliazucchi, Roseman et al., Current Biology 2016); the intensity of ego dissolution correlates with DMN disintegration (Lebedev et al., Human Brain Mapping, 2015).

The reports themselves are now standardized. The Ego Dissolution Inventory (Nour, Evans, Nutt, Carhart-Harris, 2016) is an eight-item validated instrument with items like “I felt at one with the universe” and “I lost the sense of where my body was located in space.” The most extreme agent is 5-MeO-DMT (Davis, So, Lancelotta, Barsuglia, Griffiths, 2018): the experience lasts 15–30 minutes, the dissolution is essentially complete, and the subject remembers it afterwards. This last clause is doing more work than it looks. Dissolution that wasn’t witnessed and remembered would be indistinguishable from simply not existing. Subjects come back with reports.

These results have started moving into clinical practice — psilocybin for treatment-resistant depression (Carhart-Harris et al., Lancet Psychiatry 2016; comparison with escitalopram in NEJM 2021), for major depressive disorder (Davis et al., JAMA Psychiatry 2021), for cancer-related existential distress (Griffiths et al. 2016; Ross et al. 2016). The point isn’t the therapeutics; it’s the invariant. Across thousands of recorded ego-dissolution experiences, the same structural feature appears: the experiencer dissolves, and the dissolution gets reported. Something keeps taking notes.

Meditation cessation gives us the same structure from a completely different angle. Theravada Buddhism describes nirodha-samapatti, “the attainment of cessation,” in which conscious experience is reported to stop for a period. The state is described in detail in the Visuddhimagga (Buddhaghosa, 5th century CE) and is distinguished from sleep by its preconditions: active intention to enter, active intention to exit at a pre-set time, body remaining in the entry posture throughout. Practitioners report “no time elapsed” on emergence. The briefer cousin, fruition (phala-samapatti), shows up across the Pali commentary tradition and is described in plain modern English by Daniel Ingram in Mastering the Core Teachings of the Buddha (2008) — Ingram is useful precisely because he isn’t packaging the experience in mystical language. Mahasi-style noting practice (Mahasi Sayadaw, mid-20th century Burma) reports the moment-to-moment construction of the felt self being directly observable as it assembles and dissolves.

The empirical work is recent and still building. Matthew Sacchet’s Meditation Research Program at MGH/Harvard has been documenting advanced meditative states neurally. Willoughby Britton’s Varieties of Contemplative Experience project at Brown documents both attainments and adverse effects, which matters because it’s a check against romantic interpretations. Daniel Ingram’s Emergent Phenomenology Research Consortium is a practitioner-researcher collaboration trying to make the cessations empirically tractable. The data is thinner than the psychedelic work, but the shape of the reports is the same. The practitioner emerges in the entry posture, with the timed exit honored, and (sometimes) with retrospective reports of having been “to” the cessation. The experiencer dissolved; the experiencing system continued. Same shape as the psychedelic case, produced by the opposite cause — relaxation of priors via a chemical agent versus active cultivation across decades.

The third case is pathological — dissociation and depersonalization. Depersonalization-Derealization Disorder (DDD) involves persistent or recurrent feelings of detachment from oneself or from reality. Patients describe observing themselves from the outside, feeling like a robot, watching a film of their own life. The clinical disorder is rare (~1–2% lifetime prevalence) but transient versions are very common — about half the population reports them at some point, especially under fatigue, trauma, or extreme stress. Acute drug-induced versions are well-documented: ketamine, dextromethorphan, salvia divinorum produce dissociative states reliably, and high-dose cannabis sometimes triggers depersonalization in vulnerable people. The most striking single demonstration is electrical: Blanke et al. (Nature, 2002) showed that stimulating the right temporoparietal junction during epilepsy surgery induced out-of-body experiences in patients on the operating table. The felt fact “I am located here in this body” is a constructed feature of experience, and a 5-volt pulse is enough to disrupt the construction.

In every case the same invariant survives. The experiencer goes offline in three structurally different ways — pharmacologically forced, deliberately cultivated, pathologically intruding. Something else does not. Something registers the dissolution, remembers it afterwards, makes the report. Whatever that something is, it is not the experiencer. The experiencer is precisely what got eliminated. Whatever survives is the system that produces experience — an operation, not an entity. The producing-operation persists when the produced-feeling-of-being-someone fails.

What is left to eliminate from selfhood, then? We could still say: “Fine, the experiencer dissolved, but the experiencing itself persisted. The bare fact that there is something it is like to be me right now — that’s the self. The first-person perspective. The view-from-here. Even when everything else about me is stripped, this remains.”

It is the strongest line of defense the self has. Worth taking seriously, and worth attacking.

The position has been stated more carefully than anywhere else by Thomas Nagel in “What Is It Like to Be a Bat?” (Philosophical Review, 1974). Even if you knew everything about bat neurology, echolocation, and behavior, you still wouldn’t know what it is like to be a bat. Subjective character can’t be captured by objective description. David Chalmers (“Facing Up to the Problem of Consciousness,” 1995; The Conscious Mind, 1996) sharpened the move into the “hard problem”: the easy problems explain behavior, function, access to internal states; the hard problem is explaining why there is anything it is like to undergo any of it. The first-person perspective is precisely what Chalmers thinks can’t be reduced. This is the position we have to push against.

There are three angles of attack worth trying.

The first is introspective. David Hume, A Treatise of Human Nature (1739), Book 1, Part 4, Section 6 (“Of Personal Identity”): “When I enter most intimately into what I call myself, I always stumble on some particular perception or other, of heat or cold, light or shade, love or hatred, pain or pleasure. I never can catch myself at any time without a perception, and never can observe any thing but the perception.” Hume goes looking for the experiencer and finds only experiences. The conclusion he draws is the bundle theory: there is a stream of perceptions but no perceiver-as-entity owning them. Hume’s failure to find himself has been replicated by trained introspectors — contemplatives, phenomenologists — for centuries since. The empirical observation that the experiencer doesn’t show up to introspection is robust across traditions; the disagreement is about what the absence means.

The second angle is structural. Ludwig Wittgenstein, Tractatus Logico-Philosophicus (1921), 5.62–5.641: “The world is my world: this is manifest in the fact that the limits of language … mean the limits of my world.” (5.62) “There is no such thing as the subject that thinks or entertains ideas.” (5.631) “The subject does not belong to the world: rather, it is a limit of the world.” (5.632) Wittgenstein’s point is sharper than Hume’s. The “I” in “I see” isn’t a thing inside the world — it’s the boundary of the world, the edge of the field. You can’t point to it the way you point to objects, because the pointing happens from inside it. The subject isn’t missing from the field; it’s the shape of the field.

The third angle is contemplative. The Theravada doctrine of anatta — no-self — treats the absence of an underlying self as an empirical claim, verifiable through trained introspection. The five aggregates (skandhas: form, sensation, perception, mental formations, consciousness) all arise dependently and are impermanent; no aggregate is the self; nothing underlies the aggregates. The investigation is supposed to be done directly in meditation, not by argument. Modern lay description: Sam Harris, Waking Up (2014), coming at the practice from a neuroscience-and-skepticism background. The practical exercise reduces to one instruction — look for the one who is looking. The looker never shows up. Only the looking does. Same result Hume reported, produced by a different method, in a different culture, twenty-three centuries earlier.

Three angles, three failures to locate the first-person perspective as a thing. None of them, individually, is decisive — Hume can be answered, Wittgenstein can be reread, the contemplative report can be dismissed. But the convergence is informative. Three traditions trying to find the perspective, by methods designed not to find each other, all return the same finding: the perspective doesn’t show up where it should be, given that we’re using it constantly.

Someone might still want to push: “Fine, but couldn’t I just step outside the first-person perspective and look at it directly?” Try it. Notice what the sentence requires. Stepping outside means moving from one location to another, and looking at it requires a perspective doing the looking. The proposal asks the perspective to relocate and inspect itself. The relocating and the inspecting are perspective. The very attempt to eliminate the first-person perspective deploys the first-person perspective. Every elimination move in this whole post — every layer stripped, every counter-example considered, every line you’ve read — has been performed from within a first-person perspective. We can’t get behind it. Every attempt to eliminate it deploys it.

So the situation is one of two things. Either the first-person perspective is a residue — the one thing we can’t strip, the irreducible kernel of self — or it isn’t a thing at all, and what looks like a residue is actually the act of stripping showing up to itself. The first reading reinstates the self as a thing of a special kind. The second reading says the self was never a thing in the first place; what we’ve been calling the first-person perspective is what an elimination operation looks like from the inside, while it’s running.

I prefer the second reading, but I won’t pretend the first is dead. The recursion here is genuine and it doesn’t dissolve cleanly.

Let us now look at what survived.

For every layer stripped, the same structural feature appeared: what got eliminated was a thing. Possessions are things. Roles are things. Memories, personality traits, body parts, behaviors, narrative threads, embodiment, self-models, experiencers — all things. And every thing turned out to be removable. By the structure of the argument, what’s left can’t be a thing. A thing-like residue would have failed under one of the layers; that’s what the layers did.

What it seems to be, instead, is an operation. Specifically, the operation of distinguishing self from not-self. This is the same operation bacteria perform molecularly with CRISPR-Cas, that plants perform biochemically at their root tips, that cells perform via autopoiesis, that brains perform via self-modeling — and that you, reading this now, are doing right now. None of these have a self as a separable entity; all of them have the operation. The operation is what runs at every level of biology, and the appearance of self-as-thing is what running the operation generates from the inside.

An operation requires something to operate on. There’s no distinguishing without two sides to draw between — distinguishing means boundary-drawing. But the substrate and the operation aren’t separable: you can’t have boundary-drawing without sides, and you can’t have sides without boundary-drawing. The self is neither the substrate nor the operation; it’s the operation-on-substrate where neither half could be removed without removing the other. And the operation acts on itself — distinguishing self from not-self requires the act to define what counts as self, which is what the act is doing. It’s not a paradox. It’s the basic structure of self-reference, and you’ve been performing it your whole life without ever thinking about it that way.

All of us wake, experience life, sleep and rinse and repeat ad infinitum. What we are is what we are doing — a verb, not a noun.

But I notice something interesting now. We’ve used elimination twice — once on intelligence, now on the self. Both times what survives isn’t a thing but an operation, recursive and self-applying. What if we could write that operation down — mathematically, computationally — and see what happens?