Crossing the Threshold: Everettian Emergence and the Born Rule

Abstract

Lela (2026) proves a conditional uniqueness theorem: on robust record sectors, the quadratic assignment is the only non-negative refinement-stable induced weight, given two structural conditions — internal equivalence of refinement profiles, and admissible binary saturation — with additivity carried by disjoint continuation bundles rather than the projector lattice. Lela leaves open whether physical systems meet the saturation threshold, and offers no epistemic reading of the weight. We show that Everettian quantum mechanics crosses the threshold. Saturation is established constructively: decoherence-stabilized record sectors in the universal Hilbert space admit non-demolition refining processes realizing every norm-compatible profile (Theorem 1), answering Lela’s open question. We then argue that the theorem’s remaining condition is not a posit but a consequence of Everettian emergence: decoherence constitutes the dynamical autonomy of record structure, and a weak epistemic principle — rational self-locating weight cannot track features screened from all possible evidence — converts autonomy into profile-supervenience. The relocated non-contextuality that a modal reading of saturation requires is faced openly and discharged by an Everett-exclusive argument: choices among refining contexts can be delegated to branching processes, rendering every context actual in a sibling decision-branch, so cross-context consistency is diachronic coherence among actualities. The result closes the loop on Everett’s original 1956 derivation of the squared-amplitude measure, supplying the justifications he stipulated. A final ledger itemizes what is still assumed: strictly weaker posits than those of our companion analysis, with the exponent now purchased by theorem.

1. Introduction

In 1956 Everett derived the squared-amplitude measure from two stipulations: that the measure of a branch at one time equal the sum of the measures of its successors, and that the measure be a function of branch amplitude alone. The derivation was sound; the stipulations were the problem, and seventy years of criticism — from DeWitt’s reviewers to Barrett’s reconstruction — has consisted largely of observing that they were stipulations. The decision-theoretic program of Deutsch and Wallace, the envariance argument of Zurek, and the self-location argument of Sebens and Carroll are best read as attempts to earn what Everett assumed, and the critical literature — Albert, Kent, Baker, Dawid and Thébault, Dawid and Friederich — as demonstrations that each attempt buys the conclusion with premises of comparable strength differently disguised.

In a companion paper (BMSL) we argued that the honest position is a conditional reduction: the Born rule in Everett rests on exactly two posits — a physical-measure claim (squared norm is the weight) and an epistemic bridge (rational self-locating credence tracks the weight) — and that the literature’s derivations relocate rather than discharge them. The present paper reports a development that changes the bookkeeping. Lela (2026) has proven a structural uniqueness theorem whose form is, in retrospect, a rigorization of Everett’s original argument: on robust record sectors, with additivity carried by disjoint continuation bundles, the quadratic assignment is the unique non-negative refinement-stable induced weight — conditional on two explicitly stated structural conditions. The theorem is interpretation-neutral and deliberately silent on whether physical systems satisfy its conditions and on why anyone’s credence should care about its weight.

Our thesis is that Everettian quantum mechanics, and perhaps only Everettian quantum mechanics, crosses Lela’s threshold — the qualification resting on two features we argue are Everett-exclusive: the universal-state dissolution of the measurement-enlargement problem (§3) and the choice-as-branching route to cross-context identity (§8). Crossing the threshold converts BMSL’s first posit from an assumption into a theorem-mediated consequence of emergence, while shrinking the second to its minimal core. The argument has a technical half and a philosophical half. The technical half (Sections 3–4) establishes admissible binary saturation for decoherence-stabilized record sectors in the universal Hilbert space, by a construction whose physics is deliberately familiar — the tunable quantum coin of the Everettian literature — and whose contribution is the clause-by-clause verification against Lela’s functional definitions, answering the question he explicitly leaves open. The philosophical half (Sections 5–8) begins with a confession: the saturation theorem has force only under a modal reading of “admissible,” and the modal reading relocates Gleason’s contested non-contextuality premise rather than eliminating it. We decompose the relocated debt, show that two of its three components are respectively classical-grade and theorem-grade, and devote the crux of the paper to the third: arguing that internal equivalence — the claim that weight sees only internally accessible refinement structure — follows from what Everettian emergence is, given one weak and independently motivated epistemic principle, and that the final cross-context residue admits an argument unavailable to any one-world theory, because in Everett the contexts are all actual.

Throughout, we hold ourselves to the accounting standard of the companion paper. Every premise is named when introduced and collected in a closing ledger (Section 12). The conclusion is conditional and the conditions are individually attackable; the reader who rejects the conclusion should, by the end, be able to say exactly which line item she declines to pay.

2. The threshold theorem

We restate Lela’s result in the form we need; definition and theorem numbers are his. Lela (2026) is a recent preprint; we take its definitions and its uniqueness result — his Theorem 3 — as stated, and our conclusions inherit whatever standing that source ultimately earns.

A robust record sector is a closed subspace \(R\) of a Hilbert layer \(H\) satisfying three functional conditions (his Definition 1): internal discriminability (\(R\) represents record content internally readable and distinguishable from admissible alternatives), short-horizon persistence (the content survives micro-recodings and one-step admissible evolution), and refinement closure (next-step alternatives are representable by admissible orthogonal refinements). For a global state \(|\Psi\rangle\), each sector carries a state-relative continuation bundle \(C_\Psi(R)\) — the admissible forward developments in which \(R\)’s content is realized — and a non-negative, finitely additive valuation \(\mu\) on disjoint bundles induces the weight \(W_\Psi(R) := \mu(C_\Psi(R))\). An admissible orthogonal refinement \(R = R_1 \oplus R_2\) (his Definition 3) requires the \(R_i\) to be robust sectors representing mutually exclusive readable alternatives, and — clause (iii), on which much will turn — that every continuation in \(C_\Psi(R)\) fall under exactly one \(R_i\). His Lemma 1 (continuation partition) then makes \(W\) automatically additive over admissible refinements: refinement-stability is inherited, not postulated.

The theorem (his Theorem 3): if, further, (Condition 1, internal equivalence) sectors with identical admissible binary refinement profiles carry identical weight, and (Condition 2, admissible binary saturation) every pair \((r_1, r_2)\) with \(r_1^2 + r_2^2 = \lVert \Pi_R|\Psi\rangle\rVert^2\) is realized by some admissible refinement, then \(W_\Psi(R) = c\,\lVert \Pi_R|\Psi\rangle\rVert^2\) for a constant \(c \geq 0\); normalization yields the Born form. The proof is a profile reduction followed by the Pythagorean functional equation and a Cauchy-type lemma; the exponent \(2\) is contributed entirely by the Pythagorean relation among orthogonal components. A supplementary proposition weakens exact saturation to dense saturation plus continuity of the profile function.

Three features matter for what follows. First, the additive carrier: additivity lives on disjoint sets of continuations, motivated by the same exclusivity that grounds classical probability on mutually exclusive events — no lattice structure, no betting behaviour, no symmetry postulate. The philosophically contested content of Gleason-type routes has not vanished, but it has migrated into Conditions 1 and 2, which is precisely where this paper operates. Second, Lela’s own flags: whether physically relevant systems in dimension greater than two satisfy saturation is, he writes, an open structural question, with decoherence-stabilized pointer sectors the natural candidates; and the theorem performs no ontological or epistemic interpretation of its weight. Third, the genealogy. Everett’s 1956 derivation demanded that the measure of a trajectory equal the sum of the measures of its branches and that it depend on branch amplitude alone; diachronic additivity over continuations plus amplitude-dependence is, in modern dress, bundle-additivity plus internal equivalence. Lela’s theorem rigorizes the uniqueness half of Everett’s argument. What was never supplied — by Everett or, we will argue, by his successors except at equivalent cost elsewhere — is the justification of the premises. Supplying it is this paper’s business.

3. The Everettian record layer and one structural assumption

We now specialize: the record layer is the universal Hilbert space \(H\), the state \(|\Psi\rangle\) the universal state, and record sectors are macroscopic — \(R\) is spanned by all microstates compatible with some definite, decoherence-stabilized record content: a pointer reading, a memory configuration, an environmental imprint.

Two consequences of the specialization do real work. There is no enlargement problem: the textbook realization of a generalized measurement adjoins an ancilla and dilates to a larger space, and under an operational reading of Lela’s framework this obstructs saturation — a refinement obtained by dilation refines a lifted sector in \(H \otimes H_{\mathrm{anc}}\), not the sector \(R\) that Condition 2 quantifies over. In the universal layer the obstruction dissolves, because every ancilla, apparatus, and environmental degree of freedom is already inside \(H\): the dilation space is not adjoined but occupied, and what follows is not an embedding argument but a description of physically available internal processes. This advantage is exclusive to readings on which \(H\) is universal; no operational construal of the theorem can claim it. And macroscopic sectors are never too thin to refine: record content constrains a coarse macrovariable and leaves essentially everything else free.

The little the construction needs beyond this we package as a named premise.

Ancilla Availability (AA). Every decoherence-stabilized record sector \(R\) contains an ancilla: a subsystem \(A\) with orthogonal pointer states \(|a_0\rangle, |a_1\rangle\) such that (i) the record content defining \(R\) is insensitive to \(A\)’s pointer value — \(\Pi_R\) commutes with every operator on \(A\); (ii) in the component \(|\varphi_R\rangle := \Pi_R|\Psi\rangle\), \(A\) stands in the ready state \(|a_0\rangle\), uncorrelated with the remaining degrees of freedom; (iii) \(A\)’s pointer basis is environmentally monitored, so pointer superpositions decohere on timescales short relative to the record’s persistence horizon, with redundant imprinting.

AA is a substantive premise and is priced as one (Section 12), but it is mild in the way a premise about macroscopic physics should be: it says that inside any macroscopic record there is room for one more stable bit the record does not yet constrain. A world where AA failed would be one in which some record sector had exhausted the universe’s capacity for further stable differentiation — no spin to flip, no mode to populate, no register to write — which is not our world and not any world in which experimental physics is possible. One honest qualification: clause (ii) is state-relative — a claim about the actual universal state restricted to \(R\), not about all states — and it can fail for contrived \(|\Psi\rangle\). Our results are stated for sectors satisfying AA; Section 10 notes that the credence application only ever invokes observer-containing sectors, for which AA is uncontentious.

4. The Saturation Theorem

Lela’s Condition 2 demands that every norm-compatible binary profile be realized by an admissible refinement, and he is explicit that whether physical systems meet this threshold above dimension two is open, with decoherence-stabilized sectors the natural candidates. This section answers the question affirmatively for the Everettian layer. The construction is deliberately humble — the tunable quantum coin that appears in Deutsch’s equal-amplitude decompositions, Zurek’s fine-grainings, and Sebens and Carroll’s branch-splittings. The contribution is not the device but the verification, clause by clause, that its output satisfies Lela’s definitions: that verification converts a familiar Everettian manoeuvre into the discharge of a published theorem’s premise, and clause (iii) of his Definition 3 will exact a philosophical price (Section 5) that casual uses of the device have never had to pay.

Lemma 1 (Tunable non-demolition refinement). Let \(R\) satisfy AA with \(|\varphi_R\rangle \neq 0\). For every \(\theta \in [0, \pi/2]\) there exist a unitary \(U_\theta\) and subspaces \(R_1(\theta), R_2(\theta) \subseteq R\) such that: (i) \(U_\theta\) acts only on \(A\), hence \([U_\theta, \Pi_R] = 0\) and \(U_\theta\) preserves every record sector whose content is \(A\)-insensitive; (ii) writing \(|\Psi'\rangle = U_\theta|\Psi\rangle\), we have \(R = R_1(\theta) \oplus R_2(\theta)\) with \(R_i(\theta)\) the sub-sector of \(R\) carrying \(A\)-pointer value \(a_{i-1}\), and \(\lVert \Pi_{R_1(\theta)}|\Psi'\rangle\rVert = \cos\theta \cdot \lVert\varphi_R\rVert\), \(\lVert \Pi_{R_2(\theta)}|\Psi'\rangle\rVert = \sin\theta \cdot \lVert\varphi_R\rVert\); (iii) after the monitoring of AA(iii), the pointer alternatives within \(R\) are decoherent: interference between the two components of \(|\Psi'\rangle\) has no record-functional effect over the persistence horizon.

Proof. By AA(ii), \(|\varphi_R\rangle = |a_0\rangle \otimes |\chi\rangle\) for some \(|\chi\rangle\) in the complementary factors of \(R\)’s support. Let \(U_\theta\) rotate \(A\): \(U_\theta|a_0\rangle = \cos\theta|a_0\rangle + \sin\theta|a_1\rangle\) (identity elsewhere). Then \(U_\theta|\varphi_R\rangle = (\cos\theta|a_0\rangle + \sin\theta|a_1\rangle) \otimes |\chi\rangle\). Set \(R_i(\theta) := (\text{pointer eigenspace of } a_{i-1}) \cap R\); these are orthogonal and their direct sum exhausts \(R\) restricted to the pointer span, which suffices since the construction iterates within either summand. Claim (i): \(U_\theta\) is local to \(A\), and \(\Pi_R\) commutes with \(A\)-operators by AA(i); in particular \(\Pi_R|\Psi'\rangle = U_\theta \Pi_R|\Psi\rangle\), so \(\lVert\Pi_R|\Psi'\rangle\rVert = \lVert\varphi_R\rVert\) — the refinement is non-demolition with respect to the coarse record. Claim (ii) is read off the rotated component. Claim (iii) is AA(iii). ∎

The map \(\theta \mapsto (\cos\theta, \sin\theta)\cdot\lVert\varphi_R\rVert\) is a continuous surjection onto the full profile set, so every norm-compatible profile is realized exactly; should implementability of every \(\theta\) in the continuum be doubted, Lela’s dense-saturation-plus-continuity variant stands ready, and continuity here is barely an additional assumption — realized norms vary continuously with a continuously tunable coupling.

Lemma 2 (Admissibility). Under AA, the decomposition \(R = R_1(\theta) \oplus R_2(\theta)\), taken after the monitoring of AA(iii), is an admissible orthogonal refinement of \(R\) relative to \(|\Psi'\rangle\), and each \(R_i(\theta)\) is a robust record sector.

Proof. Definition 1(i), internal discriminability: the pointer states are orthogonal and, by AA(iii), redundantly imprinted in the environment. Redundancy is the operative point. Lela glosses his conditions as functional rather than microscopic: what is required is that the record be internally readable, not actually read. Quantum Darwinism supplies exactly this — the pointer value is recoverable from many disjoint environmental fragments, so the discrimination is available to internal processes generically and cheaply. A strict reading on which readability demands actual access would disqualify Lela’s own paradigm cases (unobserved pointer sectors); we take the functional reading as intended. Definition 1(ii), persistence: pointer decoherence with environmental redundancy is the textbook mechanism of record stabilization; the identifying content of \(R_i(\theta)\) survives micro-recodings because the environment continuously re-records it. Definition 1(iii), closure: each \(R_i(\theta)\) itself satisfies AA — macroscopic sectors contain many ancillas, and consuming one leaves the supply intact — so refinement iterates. For Definition 3: clauses (i)–(ii) are immediate. Clause (iii) — every continuation of \(C_{\Psi'}(R)\) falls under exactly one \(R_i(\theta)\), none outside — holds because the pointer basis is complete on its span and decoherent: every forward development of the \(R\)-component carries a definite stable pointer record; exclusivity by orthogonality plus suppression of re-interference, exhaustiveness by completeness. ∎

Theorem 1 (Everettian binary saturation). Every decoherence-stabilized record sector satisfying AA admits, for every \((r_1, r_2)\) with \(r_1^2 + r_2^2 = \lVert\varphi_R\rVert^2\), an admissible orthogonal refinement realizing that profile. Hence the Everettian universal layer satisfies admissible binary saturation on such sectors, and — granting Condition 1, to which Sections 6–8 are devoted — Lela’s theorem applies: \(W_\Psi(R) = c\,\lVert\Pi_R|\Psi\rangle\rVert^2\).

Proof. Set \(\theta = \arccos(r_1/\lVert\varphi_R\rVert)\); apply Lemmas 1–2. ∎

Corollary. Lela’s open question — physical instances of the saturation threshold above dimension two — has an affirmative answer in the Everettian universal layer, witnessed by ordinary system–ancilla–environment dynamics.

The reader will have noticed what the proof of Theorem 1 quietly did: it chose a \(\theta\). For a different profile it would have chosen a different \(\theta\), and the processes \(U_\theta\) for distinct \(\theta\) are mutually incompatible — at most one of them is what actually happens. Saturation, as just established, is a claim about what the dynamics can do, not about what it does do — and that gap is where the paper’s philosophical cost lies. The next section takes it up.

5. What saturation costs: the modal reading

Suppose admissibility quantified only over refinements the actual forward dynamics of \(|\Psi\rangle\) enacts. Then the admissible class for a given sector contains at most one non-trivial binary profile — whichever process occurs — and Lela’s own Remark 7 shows what follows: a sparse refinement class supports only fragmentary instances of the functional equation, which do not force the quadratic form. His counterexample is explicit — for an equal-split-only class, every weight of the form \(g(s) = s^2(1 + \varepsilon \sin(4\pi \log_2 s))\) is refinement-stable — and it transfers verbatim. On the actualist reading, Theorem 1 is true but useless: it exhibits, for each profile, a possible process, while the actualist’s class collects only enacted ones. Uniqueness fails; anti-Born weights survive.

So the theorem has force only under a modal reading: a refinement is admissible if some available physical process would realize it, where “available” means implementable by unitary dynamics acting within the sector compatibly with its record structure — the precise content of Lemmas 1–2. The functional equation is then instantiated across counterfactual contexts: its two sides are evaluated in incompatible would-be histories, and it constrains a single weight function across all of them at once.

We state plainly what this is: non-contextuality, in modal dress. The demand that one fixed \(W\) satisfy the partition law for every counterfactual refinement simultaneously is the bundle-formulation analogue of demanding that a frame function assign a projector the same value in every orthogonal decomposition containing it. The route through continuation bundles did not eliminate Gleason’s contested premise; it relocated it, one level down, from a lattice axiom to a cross-context identity claim.

Saying so prepares an exact bill, for the modal reading decomposes the theorem’s premises into three items of sharply different standing. Bundle additivity is innocent: within any one context, the continuations realizing \(R_1(\theta)\) and \(R_2(\theta)\) are disjoint sets, and the weight of a disjoint union is the sum of the weights — classical-grade exclusivity, no lattice, no betting, no symmetry. Modal richness — that the counterfactual contexts exist in sufficient supply — is no longer a posit but Theorem 1: a physical fact about what unitaries, ancillas, and monitoring can jointly do inside a macroscopic sector. It could have failed: were non-demolition refinement dynamically impossible, the route would end here; that the physics cooperates is a discovery about the theory, not a stipulation within it. Cross-context identity — that it is one and the same weight function across the contexts — is where the entire relocated debt concentrates, and within Lela’s framework it is precisely the work of his Condition 1. Strip that condition away and the modal reading collapses back into the actualist one, pathologies and all.

The bookkeeping consequence shapes the rest of the paper: of three premises, one is classical-grade, one is now a theorem, and one carries everything. Sections 6–8 are therefore not one front among several but the crux of the climb. Everett, in 1956, derived the squared-amplitude measure from diachronic additivity over continuations together with the bare stipulation that the measure depend on branch structure alone; his critics have observed ever since that the stipulation is what needed earning. This section has located that old debt with new precision. The next three attempt to pay it.

6. Dynamical autonomy without probability

The foundation must be a version of branch autonomy that owes nothing to probability — otherwise the circularity charges of Baker and of Dawid and Thébault end the argument at its first step, and we state their objection at full strength. Baker: proofs of decoherence proceed via reduced density matrices and expectation values whose physical interpretation presupposes the Born rule. Dawid and Thébault: discarding interference terms because they are small presupposes that small amplitude means small significance — an implicitly probabilistic judgment — so decoherence-defined branching cannot non-circularly ground a probability derivation.

The claim we need is deliberately limited. Call it autonomy:

(AUT) After decoherence, the record-functional evolution of a sector factorizes — which discriminations are internally available, which records persist, which continuation structure obtains within \(R\) develops as it would were the companion components of \(|\Psi\rangle\), and the choice among available external contexts, absent or otherwise.

AUT concerns record-functional structure only; exact microstates are not autonomous, interference terms being small rather than zero. The claim is that their record-functional effect is null over the persistence horizon — which is all that Section 7 will need.

Following Franklin’s reply to Dawid and Thébault, AUT can be evidenced without probabilistic premises. The effects that establish emergence are dynamical: the stability of quasiclassical patterns, the persistence of records, the suppression of re-interference as a feature of the flow on Hilbert space. None of these requires expectation values to state; where textbook derivations deploy density-matrix methods, these are calculational conveniences, and the underlying claims can be cast as perturbation bounds on vectors. One premise, however, hides in “suppression” and “small,” and must be made explicit:

(NG) Dynamical relevance is norm-graded: norm-small perturbations have proportionally small record-functional effects.

Is NG the circular premise renamed? We argue not, on the ground that the norm topology is the dynamics’ own. Unitary evolution is an isometry of the norm; its continuity, its perturbation theory, and the stability results of decoherent-histories theory are all stated in it; no rival functional has any dynamical standing — there is no alternative topology in which the theory’s stability theory can so much as be formulated. Privileging the norm as the gauge of dynamical relevance is not an interpretive choice calibrated against statistics; it is reading the Schrödinger equation’s native modulus of continuity off the equation. We note with pleasure that the sharpest statement of this point in the literature belongs to a rival program: Saunders’s continuity-in-norm requirement, in his rehabilitation of branch counting, derives, as he says, from the norm on Hilbert space — we repurpose the observation and acknowledge its source.

Two clarifications complete the section. First, what is not claimed: no dynamical office is asserted for the square of the amplitude. NG concerns the norm as a grading; the exponent \(2\) enters this paper exactly once, in the Pythagorean relation inside Lela’s functional equation. The circularity that threatens dynamical defenses of \(|a|^2\) — that the square’s dynamical credentials are established through Born-presupposing statistics — is therefore not inherited here but dissolved: the squaring is done by geometry, not by physics. Second, the residual exposure: a determined critic may insist that grading record-relevance by norm is still a significance judgment. Our reply — that NG is a structural claim checkable as perturbation bounds within the formalism, whereas the circular premise is an interpretive claim about importance — is an argument, not a theorem, and the critic’s strongest rejoinder (that record-functional effect does not matter absent probability) would equally undercut every rival program. NG is logged in the ledger as challenge-shaped: the challenge being to exhibit any rival grading with dynamical standing.

7. Evidential internalism and the derivation of internal equivalence

The epistemic principle we need is weak, and its weakness is the point.

(EI) The weight that rational self-locating credence tracks cannot depend on features from which the agent’s total possible evidence is dynamically screened.

Here “total possible evidence” means the full record-functional structure internally accessible within the sector over its horizon, and the modality of “possible” is the same modality introduced in Section 5 and certified in Section 4: implementable by internal dynamics. The paper runs on one modal notion throughout; a reader who has accepted the modal reading of saturation has already accepted the modality EI quantifies over.

EI is best located by contrast with Elga’s indifference principle, of which it is the weak sibling. Elga: evidence-identical centered predicaments receive equal credence — a principle independently endorsed in formal epistemology, which in Everett yields branch-counting and perishes under refinement, as both the companion paper and Sebens and Carroll’s diagnosis agree. EI demands far less: not equal credence, but no weight difference without an evidence-structural difference. Indifference legislates the values of credence and perishes by them; supervenience legislates only the dependency base and lets a uniqueness theorem dictate the values. We are asking for less than the literature already grants elsewhere, and letting Lela’s theorem do the work indifference was too strong to survive doing.

The derivation of Lela’s Condition 1 is now a short chain. By AUT, the record-functional future of a sector is screened from companions and contexts. By Theorem 1, the admissible binary refinement profile of a sector is exactly the inventory of its possible internal evidence: every profile element is realizable as an actual record by an internal process — this is where the technical half pays epistemic rent. Hence two situations with identical profiles have identical record-functional futures: the same discriminations available, the same continuation structure, the same possible evidence trajectories. By EI, weight cannot differ between them. That is Condition 1, no longer posited: what Lela stipulates as internal equivalence, autonomy plus evidential internalism derive as profile-supervenience. A weight that distinguished profile-identical situations would track what Lela calls representational surplus — differences that make no difference to any record, hence to any possible evidence, hence to anything credence could answer to.

Two objections, faced now and revisited in Section 11. The externalist asks why weight should not track evidence-transcendent features, as objective chance is sometimes held to do. In Everett there is nothing for the transcendent feature to be: by AUT, differences beyond the profile are surplus, and a weight sensitive to surplus could never be evidentially vindicated or refuted — adopting it severs credence from confirmation entirely, gutting the confirmation structure that any empiricist account of quantum theory’s success requires. This is a strong argument and not a theorem; it is ledger item (a). The gerrymander objection asks how profiles are individuated; we inherit Lela’s individuation and note that Theorem 1 shows it maximally fine for Everettian sectors — every norm-compatible profile is realized — so nothing finer is internally accessible and anything coarser ignores accessible structure. And the reader who suspects that profile-supervenience is epistemic separability in new clothing is half right; the difference, developed in Section 11, is that it arrives here with a derivation where ESS arrived as a posit, and with its carrier fixed by a theorem where ESS’s carrier was fixed by calibration.

8. Cross-context identity: every context is somebody’s actual

Section 7 delivers that weight is a function of the profile alone. The modal residue of Section 5 remains: why one and the same function across the counterfactual refining contexts \(\theta\) — the cross-context identity that makes the functional equation bind globally rather than context by context?

The first argument is an application of what we already have. By Lemma 1(i), contexts differing in which \(U_\theta\) would be applied differ in nothing with record-functional footprint on the parent sector: the coarse record, its profile, its continuation structure are point-identical across contexts — that is what non-demolition means. A weight varying with context would vary without any profile difference, in direct violation of profile-supervenience. A reader who balks only here is balking at EI, and should be routed back to Section 7.

The second argument is the one no one-world theory can make. In Everett the agent and her choosing are physical systems, and a choice among refining contexts \(\{\theta_i\}\) can always be delegated to a quantum randomizer — an ancilla-driven branching process, available whenever AA holds, i.e. always. Under delegation the choice event is itself a branching event, and every context is enacted in some sibling decision-branch: the “counterfactual” contexts are all actual. The demand for one weight function across contexts then becomes the demand that a pre-decision agent’s weight for the parent record not depend on which of her sibling successors enacts which refinement — diachronic coherence under one’s own branching, a consistency constraint among actualities rather than a stipulation about possibilia.

Lemma 3 (Decision-branch consistency). Let a pre-decision agent face a delegated branching choice among contexts \(\{\theta_i\}\), each satisfying Lemma 1. Then profile-supervenience together with the commutation clause of Lemma 1(i) entail that the induced weight of the parent sector \(R\) is identical in every sibling decision-branch.

Proof. In each sibling branch the parent profile is preserved: by \([U_{\theta_i}, \Pi_R] = 0\), the parent norm and its admissible refinement structure are untouched by the enacted refinement. The profiles are therefore identical across siblings, and by Condition 1 — as derived in Section 7 — the weights are identical. ∎

Note what the proof does not use: at no point are the measures of the sibling decision-branches invoked — only their existence and their record structure. The argument is measure-free by construction — the point a circularity check will probe first: a justification of the Born rule that needed Born weights over the randomizer’s outcomes would be the old circle; Lemma 3 needs only that the outcomes are there.

The bridge from delegated to deliberate choices is one more application of EI. An agent whose parent-sector weight depended on whether she chose \(\theta\) herself or let a quantum coin choose would treat the provenance of the context — an evidentially idle fact, leaving no trace in any parent-sector record — as weight-relevant. Provenance-sensitivity violates profile-supervenience exactly as context-sensitivity did. This move is kin to Wallace’s branching-indifference arguments, and the kinship should be acknowledged with the difference stated: his indifference is a rationality axiom within a decision theory; ours is a consequence of a principle already in play, purchased once in Section 7 and merely spent here.

Three scope guards. What has been earned is non-contextuality of weight assignments over record sectors, not of value assignments; Kochen–Specker contextuality of hidden values is untouched and irrelevant, Everett having no hidden values. The argument needs branching choices to be available, not universal: delegation suffices, and AA keeps a randomizer always in stock. And the staked ground should be credited precisely: Bub and Pitowsky observed that the Deutsch–Wallace rationality constraints are justifiable only on the Everett interpretation, where all outcomes occur relative to branches — the germ of the present argument, grown in decision-theoretic soil; and Khawaja, defending branch-counting against the Born rule, makes the corresponding identification — that in Everettian terms noncontextuality amounts to branch-dependence — and flags it as exactly what the probability-objectors reject. Neither develops the choice-as-branching-event argument; Wallace’s noncontextual-inference theorem reaches a cousin of our conclusion with rationality axioms this route does not need. Gleason had to posit non-contextuality, and hidden-variable theorists may always reject it by denying counterfactual definiteness. In Everett it is a coherence constraint among actualities: every context is somebody’s actual.

9. Assembly

Theorem 2 (Born weight at the emergence threshold). Grant: (i) non-probabilistic emergence of decoherent branching — AUT, supported by NG (Section 6); (ii) evidential internalism — EI (Section 7); (iii) that the agent’s self-locating credence over admissible record alternatives exists and is finitely additive over exclusive alternatives. Then for every decoherence-stabilized, AA-satisfying record sector \(R\), rational self-locating credence is

\[ C(R) = \frac{\lVert\Pi_R|\Psi\rangle\rVert^2}{\lVert\Psi\rVert^2}. \]

Proof. Sections 6–8 derive Lela’s Condition 1 from (i) and (ii), with Lemma 3 discharging the cross-context identity that the modal reading of Condition 2 requires; Theorem 1 establishes Condition 2 for AA-sectors. Premise (iii) makes credence a non-negative finitely additive valuation on disjoint admissible continuation bundles — an extensive bundle valuation in Lela’s sense — whose induced sector weight, satisfying Conditions 1 and 2, is by his theorem \(c\,\lVert\Pi_R|\Psi\rangle\rVert^2\). Normalizing over a complete admissible decomposition fixes \(c = \lVert\Psi\rVert^{-2}\). ∎

The logical form deserves emphasis: conditional, with conditions individually motivated and individually attackable — the methodology of the companion paper, retained. What has changed is what the conditions are.

10. The slimmed bridge

The companion paper’s epistemic bridge, MCSL, asked the agent to align credence with the objective measure — which obliged the companion paper to defend squared norm’s claim to be the measure, the burden its Section 3.6 carried uneasily. The present route asks for less, and the difference is the paper’s payoff. The bridge now consists of premise (iii) of Theorem 2 alone: self-locating credence exists, and is additive over exclusive admissible alternatives. Nothing in the bridge selects the quadratic form; selection is performed by the uniqueness theorem, operating on conditions that Sections 4 and 6–8 derived from the physics and one weak epistemic principle. The companion paper’s Section 14.2 objection — why this measure rather than another? — thereby receives a structural answer: there is no other for credence to be, once credence supervenes on what evidence could ever show and the dynamics supplies refinements in every proportion.

Two inheritances are recorded rather than re-litigated. The question whether self-locating credence in a branching universe is well-posed at all — whether there is a coherent epistemic predicament between branching and observation for credence to address — is the companion paper’s Section 14.6, defended there via the distinction between indexical and prospective registers; it enters here as premise (iii) and is priced in the ledger as item (c). And the restriction to AA-satisfying sectors costs the credence application nothing: an agent’s predicament always concerns observer-containing sectors — sectors compatible with her own records existing — and such sectors satisfy AA on any non-contrived universal state, containing as they do the most record-saturated macroscopic structure there is.

11. Stress tests

11.1 The Dawid–Friederich transfer. Their case against Sebens and Carroll has a precise anatomy: a plausible general principle (epistemic separability), a specialization to quantum mechanics requiring a choice of formal carrier for the local epistemic situation, and the demonstration that the chosen carrier — the reduced density matrix — has no motivation independent of the empirical success of Born-rule physics. Calibration, not derivation. Run against the present route, the transfer fails at the specialization step, and it is worth saying exactly why. Our general principle, EI, is independently motivated, non-quantum, and strictly weaker than a principle (Elga’s) the formal-epistemology literature already endorses. Our specialization — record-functional structure is the admissible refinement profile, and the realizable profiles are exactly the norm-compatible ones — is mediated by Theorem 1, a result proven from unitarity, Hilbert geometry, and AA, not read off statistically vindicated practice. The reduced density matrix wears Born weights on its diagonal; the profile set wears only what unitary couplings can do. The hostile reading — epistemic separability in new clothing — is therefore half right, and the half matters: this is separability with a derivation, where ESS was separability as posit; the clothing is a theorem.

11.2 The circularity charges. Section 6 met these in full; what remains is the residue. Against Baker: no reduced density matrices or expectation values appear in AUT’s support — the emergence claims it rests on are cast as perturbation bounds on vectors, not as facts about expectation values. Against Dawid and Thébault: NG grades record-relevance by the norm, which is a structural claim checkable as perturbation bounds within the formalism, not the significance judgment their objection targets. The standing exposure is the challenge to exhibit a rival grading of dynamical relevance with any dynamical standing; until one is produced, NG stands as the only candidate the theory itself offers, and the ledger records it as challenge-shaped.

11.3 Saunders’s rehabilitated branch-counting. The most serious rival route, and partly a creditor: his counting rule achieves convention-freedom by a continuity-in-norm requirement, and we have repurposed his observation that this requirement derives from the Hilbert norm itself. The difference is bookkeeping. His route lets the measure enter through the topology that individuates the branches to be counted, implicitly; ours names the norm-gradedness premise, prices it, and lets the quadratic form be earned by theorem rather than by count. Whether his implicit dependence is lighter or heavier than our explicit one is a fair question for a referee; we contend that an itemized bill beats an unitemized one at equal cost.

11.4 Khawaja’s indexed branch-counting. The sharpest standing denial of our conclusion is a rival rule, not a circularity charge: Khawaja’s indexed branch-counting (IBC), a well-defined self-locating credence that diverges from Born, defended positively on accuracy-theoretic grounds and from a requirement of exchangeability over event-sequences. It defuses the cheap stop — that branches are too vague to count — so we cannot lean on incoherence, and do not; and it attacks the very Saunders proposal §11.3 credits. That last is easily discharged: §11.3 borrows only Saunders’s observation that continuity-in-norm derives from the Hilbert norm, never his counting rule, so an attack on the rule leaves our use of him intact. Indeed Khawaja and we agree that Saunders’s count does not by itself earn Born; we part on the moral — he takes the honest count to diverge from Born, we take the quadratic weight to be forced by Lela’s theorem, with counting nowhere in the derivation.

What IBC cannot evade is the uniqueness theorem. A non-Born refinement-stable induced weight is impossible, so its divergence from Born requires it to fail refinement-stability or internal equivalence — and Theorem 1 shows how. The saturation construction splits any sector, by a norm-fixing non-demolition coupling, into as many equal-norm sub-branches as one likes: the count is unbounded and refinement-relative exactly where the norm is invariant. Under Section 5’s modal reading every such individuation is available, and under Section 8’s delegation each is enacted in some sibling; there is no privileged count for IBC to index. What it fixes by fiat, our framework reads as representational surplus — a choice that leaves no record, hence by evidential internalism is weight-irrelevant. Its exchangeability premise is, structurally, the indifference principle Section 7 diagnoses as perishing under refinement, now backed by an accuracy-theoretic warrant rather than a bare symmetry; the warrant is the novelty, and it is exactly a positive case for tracking what EI forbids.

We claim no knockout. If that accuracy case makes the index genuinely evidence-relevant, EI is false and IBC stands — which only repeats that everything turns on ledger item (a). What we add is a burden IBC has yet to discharge and squared norm discharges trivially: a credence diverging from Born predicts non-Born relative frequencies as typical, and the statistics an embedded experimenter records are Born. Exchangeability governs how a rule treats reorderings of a fixed frequency; it does not, by itself, say why the recorded frequencies are the Born ones. Until it does, IBC buys its rejection of EI at the price of the empirical contact our route keeps for free.

11.5 The ad-hocness probe. The suspicion that the package was reverse-engineered to yield Born is answered by its exposure to failure. If refining couplings necessarily disturbed parent records, Theorem 1 fails; if decoherence did not screen contexts, Section 8’s first argument fails; if dynamical relevance were not norm-graded, Section 6 fails; if self-locating credence is ill-posed, premise (iii) fails. Each component can lose on structural or physical grounds, and none is tunable to save the conclusion — four independent risks, where a reverse-engineered package would take none.

12. The ledger

The companion paper closed by pricing two posits. This paper closes by showing the revised bill — all of it, including the small print.

# Premise Where Status
a Evidential internalism (EI) §7, §11.4 Open posit; Khawaja’s indexed branch-counting (§11.4) is the worked-out rejection. Refusing EI severs credence from confirmation (§7) — strong reply, not theorem.
b Norm-gradedness (NG) §6 Open posit, challenge-shaped: no rival grading has dynamical standing.
c Well-posed additive self-locating credence §9(iii), §10 Inherited from companion §14.6; untouched here.
d Ancilla Availability (AA) §3 Local; mild for macroscopic sectors; state-relativity priced, defused for observer-containing sectors (§10).
e Functional reading of “readable” §4, Lemma 2 Interpretive; strict reading would disqualify Lela’s own paradigm cases.
f Delegation-availability §8 Follows from AA; listed for completeness.

Items d–f are local and, we have argued, defused where they arise; a referee is nevertheless entitled to the full count, and the count is six, not two. The headline comparison is with the companion paper’s two posits: squared norm is the objective measure and credence tracks the measure. These are replaced by (a) and (c) — credence answers to possible evidence, and credence exists and adds — which are strictly weaker, individually motivated, and jointly silent on any exponent. The exponent 2 is purchased nowhere in the ledger; it is purchased by the Pythagorean relation inside a uniqueness theorem, which is to say by geometry. Everett stipulated in 1956 that the measure of a trajectory be the sum of the measures of its branches and a function of their amplitudes alone, and was told for seventy years that the stipulations were the problem. They were. We have tried to show that, in the one interpretation where every outcome is somebody’s actual, they were also payable.

References