A small bell rings in a quiet room. The sound begins sharply, opens into a shimmer, then disappears. You can point to the bell, the vibrating air and the pressure wave reaching the ear. You can also notice the heard tone itself. Now make one further move: locate the one who heard it.

Perhaps attention moves behind the eyes. There is pressure in the forehead, a faint outline of the head, an inwardly spoken word such as “me”, or a sense of being positioned at the centre of a visual field. Each candidate seems promising for a fraction of a second. Then the difficulty appears. The pressure is felt. The head-image is seen inwardly. The word is heard. The centre is spatially represented. Every located observer has crossed the subject-object seam and become another item in the scene.

The usual intuition says that a hidden witness must remain just behind the latest object. That intuition contains an unnoticed assumption: anything real must be findable in the same manner as the things it knows. The search therefore keeps installing a smaller observer behind the previous observer. Philosophers know this as a regress problem. Contemplative traditions turn it into a practical instrument.

The central question is not whether a subject exists. It is more precise: what happens when a method asks the apparent knower to present itself as an object, then refuses every object as the final knower? That question is shared locally by witness practice, Ramana Maharshi’s self-inquiry, Christian apophatic method, Mahāmudrā and Dzogchen. Their answers are not shared.

flowchart TD
  A["The listener behind the bell"] --> B["How the seam folds"]
  B --> C["Five routes, five stopping rules"]
  C --> D["Same report, different causes"]
  D -. "re-test" .-> A
  classDef first fill:#0a1424,stroke:#D95545,stroke-width:2px,color:#e6f3ff
  classDef step fill:#0a1424,stroke:#3f7da0,stroke-width:1px,color:#e6f3ff
  class A first
  class B,C,D step

“Attention attending to itself” sounds like a single operation. It hides at least three. A practitioner may monitor where attention is allocated. They may notice a bodily sense of being the observer. They may also question whether any noticed item can be the knowing subject. Only the third move creates the puzzle at issue here.

I will call the shared local mechanism reflexive de-objectification. This is a practitioner synthesis, not a term claimed by the five traditions. It names an operational pattern while leaving their doctrines intact.

The third movement matters most. Without it, the practice produces a sophisticated homunculus: a calm, detached witness watching the rest of life. That stance can be useful for emotional decentring. It can also become another identity. Witness traditions themselves often distinguish the provisional observer from the final claim. Buddhist nondual traditions explicitly challenge grasping at a watcher.

A small formal constraint

Formal notation can compress the logic, provided it does not pretend to explain awareness. Let C_t be the contents explicitly available at time t. Let F_t(x) be an operation that makes some candidate x explicit, perhaps by attention, imagery or verbal report. Let K_t name whatever condition enables those contents to be known. Then:

Here s means a candidate subject, such as a head-centred sensation. The equation says only that an explicit representation is not automatically identical to the whole enabling condition. It assumes a distinction between content and enabling relation. Remove that assumption and the notation loses its ontological force. Nondual traditions may deny that knower and known are ultimately two. The formula therefore regulates inference; it does not decide metaphysics.

The five routes resemble one another most strongly in the middle of the operation, not at the beginning or end. Each interrupts ordinary object-directed cognition. Each refuses to identify the final knower with an item that appears. After that local convergence, their paths separate.

A disciplined comparison therefore asks five questions. What initiates the turn? Which candidate is cancelled? What is said to remain? What kind of warrant supports that claim? Where does the comparison break? The stopping rule is as important as the shared manoeuvre.

Witness practice: the observer is provisional

“Be the witness” can mean many things, from ordinary emotional distancing to a precise Advaita discrimination. Śaṅkara’s position is more demanding than installing a calm inner spectator. At the empirical level, consciousness can be described as the witnessing presence, sākṣin, in relation to mental states. Yet the absolute account removes even relational properties such as being a witness of something. The self is not a second entity inside the mind and not a metacognitive module watching other modules. It is claimed to be self-illuminating and non-objectifiable. (Dalal, 2021)

The local operation is clear. Body, sensation, thought and the felt witness can each be known, so none is accepted as the final self. The strong Advaita conclusion is more than “the observer is hard to find”. It is that awareness is the condition of appearances and, ultimately, nondual Brahman. That conclusion belongs to a philosophical and soteriological argument extending far beyond the local exercise.

Self-inquiry: trace the claimant, not the answer

Ramana Maharshi’s short instructions begin when a thought arises. The practitioner asks, in effect, “To whom has this arisen?” The answer “to me” triggers the further question “Who am I?” The point is not to generate a verbal definition. It is to turn towards the felt claimant and follow the “I”-thought towards its source until the thought subsides. (Sri Ramanasramam, Who am I?)

This route differs from generic witness practice. It does not begin by occupying a detached position. It uses the ego’s own first-person claim as the thread of inquiry. Each proposed answer, body, biography, mood, location or subtle observer, is another objectified candidate. The practice ends, within Ramana’s Advaita framing, in the Self rather than in an indefinitely repeated question.

A psychology experiment may study changes in self-referential thought, attentional allocation or report. It cannot quietly replace “Self” with a measured variable and claim that the tradition has been verified. Nor may the tradition’s authority substitute for an intervention when a neural or behavioural mechanism is proposed.

Apophatic method: negation reaches beyond the self

Pseudo-Dionysius offers the most instructive disanalogy. His apophatic method proceeds through affirmations about God, negations of those affirmations, and a movement beyond the opposition between affirmation and denial. The intellect relinquishes sensible and intelligible predicates in relation to the divine mystery. This is not simply an inward search for a pure observer. (Pseudo-Dionysius, Mystical Theology; Corrigan & Harrington)

The local resemblance is a disciplined refusal to mistake a concept or representation for its ultimate referent. The break is decisive. The target is God beyond predication, and the practice belongs to Christian scripture, liturgy, hierarchy and a theology of return and union. Apophasis preserves a grammar of cancellation, not an identity between God, self-luminous awareness and rigpa.

This difference creates a useful test for comparisons. If removing the theological object, ecclesial context and telos leaves the proposed mapping unchanged, the mapping has probably retained only the word “negation”. That is lexical analogy, not structural preservation.

Mahāmudrā: mind looks without finding a point

Tilopa’s Ganges Mahāmudrā instructions contain the striking turn in compact form: mind looks at mind. The text does not direct the practitioner to locate a hidden core. It emphasises non-fixation, the absence of a focal point and the release of the perceiver-perceived opposition. Thoughts are not defeated by a second watcher; their arising can be recognised without grasping. (Tilopa, Ganges Mahāmudrā)

The shared cancellation is easy to see. Any point selected as “mind” becomes an appearance. The Mahāmudrā endpoint, however, is articulated through Buddhist accounts of emptiness, luminosity, non-grasping and liberation. It does not affirm an eternal substance-self. Its practical idiom also belongs to transmission, training and ethical context, not a context-free attentional trick.

A modern researcher can operationalise features such as reduced subject-object structuring, effortlessness or flexible non-grasping. These remain constructs. They do not exhaust Mahāmudrā, and a high score on a nondual-awareness questionnaire does not certify realisation. (see Dunne, 2011; Meling, 2022)

Dzogchen: recognition is not blankness

Dzogchen centres direct introduction to rigpa, followed by decisive recognition and confidence in the self-liberation of arising thoughts. Garab Dorje’s three statements organise that arc. Later instructional texts repeatedly warn against confusing recognition with an inert gap, dull calm or absence of thought. (Garab Dorje; Mipham; Deroche & Sheehy, 2022)

This warning sharpens the shared mechanism. When dull blankness appears, it too can be noticed. Mipham’s instruction turns attention towards that which is aware of the dull state, not so that another witness can be manufactured, but to distinguish ordinary mind, sem, from rigpa. Dzogchen’s tradition-internal claim concerns primordial knowing and its inseparability from appearances. Calling every quiet or observerless state “rigpa” erases the distinction the practice was designed to protect.

Mahāmudrā and Dzogchen can sound close in translation, especially around non-grasping and nature of mind. Their textual lineages, technical vocabularies and modes of introduction still matter. A responsible comparison names exactly which relation is preserved and does not declare doctrinal equivalence from similar English phrases.

stateDiagram-v2
  [*] --> Claim
  state "A small formal constraint" as Claim
  state "Witness practice: the observer<br/>is provisional" as Test
  state "Self-inquiry: trace the<br/>claimant, not the answer" as Revision
  Claim --> Test: expose mechanism
  Test --> Decision: evidence survives
  Test --> Revision: rival survives
  Revision --> Claim: narrow or rebuild
  Decision --> [*]

Imagine five people leaving five very different settings. One has completed Advaita inquiry. One has received a Mahāmudrā instruction. One has practised Christian apophatic prayer. One has endured a sleepless night. One has read a paragraph describing nondual awareness. Each says, “There was no separate observer.”

The sentence is evidence that the sentence was produced. With careful elicitation, it can become evidence about the reported structure and sequence of experience. It does not yet identify the cause. Recognition, absorption, reduced arousal, conceptual expectation, semantic compliance, dissociation and memory reconstruction can all generate overlapping language. A report becomes scientifically useful when rival generators are designed into the study rather than mentioned after the result.

A discriminating protocol

The smallest useful study should test a narrow functional hypothesis, not “whether nondual awareness is real”. One candidate is the flexibility hypothesis stated earlier: trained reflexive de-objectification should permit deliberate movement among engagement, observer stance and non-objectifying modes while preserving task performance and affective contact.

Recruit one clearly specified practitioner population and an active control group trained in ordinary attentional monitoring. Do not mix lineages and call the result “contemplative expertise”. Freeze the exact instructions. Separate a teacher’s tradition-internal judgement from the laboratory construct. Both can be recorded, but neither should silently validate the other.

Use four conditions: ordinary content focus, a detached witness instruction, reflexive candidate cancellation, and a semantic-mimic condition that supplies the expected vocabulary without the practice. Mark the moment of each claimed transition. Follow it with an interview conducted by someone blind to condition, a surprise cue-removal test, a mode-switching task, an affectively salient but ethically mild stimulus, and a measure of ordinary task control.

Pre-register three outcome families. The first is phenomenological: sequence, spatiality, effort, centre, agency and subject-object structure. The second is functional: switching cost, cue dependence, calibration and affective responsiveness. The third is safety and functioning: distress, sleep, derealisation, emotional numbing and daily impairment. Neural or physiological measures may be added, but they must predict something beyond the labels already supplied by the participant. (see Varela, 1996; Lutz & Thompson, 2003; Petitmengin, 2006)

A positive result would permit a limited conclusion: under this protocol, trained participants exhibited a reproducible functional and reported pattern that resisted specified mimics. A negative result would weaken that pattern, expose an insensitive assay, or show that the training dose and construct were mismatched. Neither result decides Brahman, God, emptiness or rigpa.

Safety is part of construct validity

An observerless or selfless report can be welcomed, frightening or mixed. A study of meditation-related challenges in selected Western Buddhist practitioners and experts identified a wide range of cognitive, perceptual, affective, somatic, motivational, self-related and social difficulties. Its purposive sample does not estimate population prevalence, but it defeats the assumption that all contemplative change is benign. (Lindahl et al., 2017)

A preregistered cross-sectional study by Pons and colleagues (2026) compared people reporting meditation-triggered depersonalisation or derealisation-like states with people reporting other triggers. The groups could show similar phenomenological features while differing in appraisal and emotional tone; distress was not absent from the meditation group. Because the sample was self-selected and cross-sectional, the study does not establish causation or diagnose contemplative states. It does show why valence, functioning and context cannot be inferred from “no self” language alone. (Pons et al., 2026)

The protocol therefore needs explicit stop rules. Sustained fear, severe sleep disruption, emotional numbing, inability to function, loss of voluntary control, or difficulty re-engaging consensual reality should pause the experiment and trigger qualified assessment. A spiritual interpretation must never block clinical support. Conversely, clinical resemblance alone should not erase a participant’s meaning or tradition. The safe research question is differential, not dismissive: which pattern is present, what is its trajectory, and what support does this person need?

type ClaimType =
  | "instruction"
  | "phenomenology"
  | "soteriology"
  | "metaphysics"
  | "mechanism";

type EvidenceLevel =
  | "text"
  | "report"
  | "interview"
  | "intervention"
  | "multimethod";

type PreservedRelation =
  | "turn"
  | "cancel"
  | "release"
  | "context"
  | "endpoint";

interface ConcordanceRecord {
  sourceRoute: string;
  claimType: ClaimType;
  targetDomain: string;
  evidenceLevel: EvidenceLevel;
  population: string;
  preservedRelations: PreservedRelation[];
  breakConditionStated: boolean;
  mappingGrade: string;
  allowedClaim: string;
  prohibitedLeap: string;
  nextTest: string;
  rivalGenerator: string;
}

// Invariants
// 1. No metaphysical identity follows from report evidence.
// 2. Machine-design analogy never establishes machine awareness.
// 3. No structural grade is issued without a break condition.
// 4. Endpoint identity needs independent argument, not shared vocabulary.

sequenceDiagram
  participant Q as Question
  participant M as Mechanism
  participant E as Evidence
  participant D as Decision
  Q->>M: The listener behind the bell
  M->>E: A small formal constraint
  E-->>M: compare rival explanations
  E->>D: Preserve the local mechanism and refuse the unearned endpoint
  D-->>Q: revise scope or act

The tempting decision is to choose between two crude positions. Either all five traditions discovered the same awareness, or their resemblance is too vague to matter. Reflexive de-objectification supports a better decision. Preserve the local mechanism that can be stated and tested. Refuse to transfer the endpoint that has not been preserved.

For comparative philosophy, this means replacing “Are they saying the same thing?” with a typed question: is the proposed relation lexical, operational, phenomenological, structural, empirical or metaphysical? A comparison earns strength by naming its break condition, not by hiding it.

For contemplative science, a first-person report should set the next intervention. It can constrain temporal order, felt spatiality, effort and self-world structure. It cannot self-certify its cause. The study must include semantic mimics, detachment, low arousal and context effects as rival generators, plus safety and functioning as primary outcomes.

For personal practice, non-finding can be an honest moment in inquiry. It is not a certificate of attainment. Dullness, fear, numbing and impaired functioning deserve attention rather than reinterpretation as advanced states. Tradition-specific guidance and qualified clinical support answer different questions and may both be needed.

For artificial systems, the implication is equally strict. A model can inspect attention allocations, describe its self-model or produce an observerless report. Those are functional achievements. Recursive self-description is not evidence that transformer attention has become awareness. The mechanism in this paper can inspire matched architectures and interventions, but any consciousness claim still requires a declared candidate boundary, theory and measurement bridge.

A five-line discipline is usually enough to stop the slide from resemblance to equivalence. Write the source claim in its own vocabulary. Name one relation the target genuinely preserves. State the first point at which the mapping fails. Build the strongest rival that could produce the same observation. End with the decision or experiment that the comparison changes. If any line is blank, lower the claim before adding more theory. This compact record also makes disagreement productive: one critic can challenge the preserved relation, another the rival generator, without pretending that the entire contemplative or scientific worldview must be accepted at once.