Ten questions a physicist asks first

“Isn’t this just decoherence plus a postulate?”

The ingredients are exactly those: unmodified quantum mechanics, the decoherence the bath already produces, and one postulate, that one record history is actual. The content is in what the postulate is attached to and what follows from it.

Decoherence explains the suppression of interference. It does not supply the occurrence of an outcome, and it is not the same thing as anchoring: a finite bath gives back what it took, a record that can still be read back is not yet a fact, and the laboratory has shown both (Brune et al. 1996; Chapman et al. 1995; Kim et al. 2000). What ACT adds to decoherence is an event law with three parts. When: at release, when the record is beyond coherent recovery by the system’s accessible future, at a rate set by irreversible record production, $h=\dot\Phi_{\rm irr}$ in the nondemolition case, with no new constant. What: the alternatives the record architecture supports, by the selection theorem. Which: Born weights. The second commitment of Paper I, an exclusive outcome register, is a theorem in the record-algebra class of Paper II, which is why one postulate is what remains within that class.

Carried by: Mechanism, the event law; ledger: the event law; exclusivity as a theorem; one history actual.

“Isn’t this Many Worlds? The equation is the same.”

Many Worlds shares ACT’s equation, and a real wave that never collapses. The two part at the outcome, and there they are opposites. In Many Worlds every outcome happens, each in a world of its own, with a copy of the observer in each. The bath only draws the borders between those worlds; decoherence was brought in later to say where the borders fall (below). In ACT the bath makes the mark, and one history of marks is actual.

No experiment decides between the two. Every detection is one mark, as both predict. ACT takes that mark at face value; Many Worlds counts every other outcome as a world of its own, which no experiment can reach. And where every outcome happens, a “50% chance” needs a new meaning; in ACT one outcome happens, with an ordinary chance.

On ACT the rest of the wave is what it was before the event: the structure of one real wave, holding possibilities for a mark, weighted by their amplitudes. Two senses of “real” are kept apart: the unselected alternatives are real as structure of the one wave — they carry amplitude, can still interfere while no record is released, and can anchor later where the record left something unrecorded — and they are not real as fact, since nothing has happened in them (The Picture). To the objection that the rest of the wave holds worlds in all but name, ACT’s answer is its one commitment. Every component carries its own records, but only one history of them is actual, and only its records are marks. The objection applies to ACT no less than to Bohm, since both leave the unselected components in place, and the answer is a commitment, defended, not derived.

What the one commitment buys is explanatory. Probability is ordinary, one outcome with a Born-weighted chance, and that is what lets the weights be forced (Objection 4); Everett’s derivations of the Born rule are still contested. Temporal order belongs to the one history, made by its events. Its empirical standing is set out under Objection 9, and whether Many Worlds is simpler under Objection 11.

Carried by: The Picture, what ACT is not; the record history; ledger: superposition; one history actual.

“And to Bohm, the collapse models, consistent histories, QBism? What is actually new here?”

The main accounts keep a particle somewhere, or they keep every outcome. Copenhagen brings the particle back at the detector, through a measurement it never defines. Bohm keeps it all along, on a path the wave steers. The collapse models squeeze the wave down to one place, with a new law and new constants. Many Worlds keeps no particle, but it keeps every outcome: a mark at every place the wave could leave one, each in a world of its own. The bath only draws the borders between those worlds. Those who say plainly that there are no particles, only fields, part ways with ACT at the detector. Hobson (2013) lets the field collapse there and leaves the cause open. Zeh (1993), who gave up the collapse as well as the particle, was left with every outcome. ACT keeps no particle, and one outcome. The electron is a wave from start to finish; the particle is the mark the bath makes, and one history of marks is actual. ACT takes the third road: no particle, no collapse, one history, and the bath makes the mark; the event ontologies below come nearest to it.

Other interpretations ask how a superposed quantum state yields a definite result. ACT begins with a real wave and asks how the environment produces a definite event. The difference is in the question, and it has two parts. First, the wave is not a superposed particle: a superposition is the structure of one wave, not a particle holding two classical properties at once, so ACT needs no further worlds, no continuously definite particle trajectory and no spontaneous collapse to say what a wave with several lobes is; what it needs in order to say which lobe is anchored is one commitment, that one record history is actual. Second, measurement is a physical process, not an act of observation: the detector couples the system to the bath, the bath keeps a record, and it is this coupling, together with the irreversible release of the record, that ACT identifies with anchoring; the Rock anchors with nothing reading it. Several of the main accounts predate decoherence theory and incorporated it afterwards, while quantum Darwinism makes it central from the outset and leaves open why one outcome is actual. ACT begins with the environment rather than adding it afterwards. The distinction lies in a physical ontology of how definite events arise from quantum waves; the decoherence phenomena themselves are the same for every account that keeps unitary quantum mechanics. The flagship’s §9 says this in a page.

Each of those also pays differently. Bohm adds particle positions and a guidance equation, and needs a preferred foliation to be made relativistic. GRW and CSL add a noise field and two constants; the noise heats matter, and the original parameter values are now excluded by the underground radiation searches. Consistent histories make facts relative to a framework and leave the choice of framework open. QBism and relational quantum mechanics decline, by design, to give any physics of measurement. ACT adds no field, constant, threshold or noise, keeps the global dynamics unmodified, and adds one primitive. The table below sets the accounts side by side on one further point: when each fixed its postulates, and what decoherence does in it.

The nearer relatives are the event ontologies, and ACT claims the lineage: Heisenberg’s potentia, Haag’s events as the transition from possibility to fact, Belavkin’s nondemolition principle, Kent’s late-time records, Fröhlich’s ETH approach, and Zhang’s event anchoring. What ACT adds is the combination. The Schrödinger equation is left exactly as it is, where ETH evolves stochastic states. The boundary is stated against the system’s accessible future rather than the full future algebra or a final hypersurface, with a graded bound and the scrambling separation. The event rate is the jump intensity of the record channels the bath physically supports, where event-enhanced quantum theory postulates a detector coupling of its own. And the selection theorem says that fragment records of a position-diagonal coupling can support a localization instrument and never a coherent momentum kick. The new core is narrow and the papers say so. Zhang’s paper leaves single-outcome event generation open as a separate problem; ACT answers it by its one postulate, with the release criterion fixing when: answered, not derived.

When each interpretation met decoherence

Copenhagen, Everett and Bohm fixed their postulates first and were told decades later what decoherence does for them. The collapse models must outrun it, and they show the pattern twice: the collapse noise was postulated without a damping partner, and the partner was added later. Decoherent histories and quantum Darwinism do put decoherence first, but as a condition on descriptions or an account of objectivity; neither has a bath that produces an event. Wherever decoherence enters one of these accounts it fixes the alternatives and produces none of them; which one occurs is supplied elsewhere, or not at all.

The nearest neighbour on this point is Fröhlich’s ETH approach (events, trees, histories), which ties events to the shrinking of what can still be measured as massless radiation escapes; the decoherence data are the same for every interpretation that keeps unitary quantum mechanics.

Carried by: Mechanism, decoherence is built in; ledger: release; the selection theorem; no new field, constant, threshold or noise; Lecture 9, the comparison table.

“The Born rule isn’t derived.”

The weights are no longer carried over from the ensemble. They are forced by things the theory already says — one wave, with one law of chance that is the same whatever the basis; one history, so the chance of “this record or that one” is the sum; sharp records, with no record where the wave has no component, joint records included; records that can always be refined — together with two requirements on the chance of a record that the theory states and does not derive: a record’s chance here does not change when a record is made elsewhere, for every division of the world into here and elsewhere (the Bell requirement is its spacelike case), and a record’s chance changes continuously with the wave. From these the weights are the squared amplitudes, by Zurek’s envariance argument in the no-signalling form Barnum gave it (the flagship, §4 and Appendix A; the proof in full, each step tagged with the premise it uses, is Appendix A of Paper II).

The argument has been contested where every outcome occurs (Schlosshauer and Fine 2005; Mohrhoff 2004; J. Zhang 2026), because there what a chance is comes first; here one record is actual and its chance is a chance. The only noncontextuality the argument uses is the one no signalling supplies, and the additivity it needs, which no account of the weights can do without, comes from one history within a record algebra and from no signalling across algebras. Each alternative rule fails a named premise: squaring the weights and renormalizing lets a distant record shift local statistics; a rule that is not the same in every basis fails the one law; the no-signalling toy model of Galley and Masanes makes no record ever certain. Bohmian mechanics has a no-signalling route of its own, in a different form: Valentini showed that |ψ|² is the only distribution of its configurations that forbids signalling, given the guidance law. The difference is where the forcing acts — on a distribution over added variables there, on the chance of a record here, because in ACT the complete state is the wave and there is nothing to average over. What is not derived is that there is chance at all, which comes with the one commitment, that one history is actual; the two stated requirements; and that records can always be refined. In the exactly solvable models the conditioned weights also form a martingale, so the single-history law and the master equation agree; that descent, which earlier versions of this page offered as the standing, is now the consistency check. An independent referee audit of the argument preceded the change (9 October 2026); earlier drafts of the program claimed a derivation from a mechanism, and those claims stay superseded.

Carried by: ledger: Born weights forced; the consistency check; Lecture 6; the flagship, Appendix A; Paper II, Appendix A.

“‘Irreversible’ is relative to what can be controlled. Echo experiments undo it. So the accessible algebra is a free parameter.”

It is relative to what is controlled, as the echo experiments show. Magnetization spread through a solid has been brought back as an echo (Rhim, Pines & Waugh 1971), and information scrambled across 53 qubits has been refocused (Mi et al. 2021). Spreading into many degrees of freedom is not by itself a record, and a dephasing an echo undoes was never one. ACT counts it as zero events, which is what the spin echo on the Evidence page is there to show.

What the accessible algebra is relative to is the physics of the coupling and the apparatus, the detector specification that quantum-trajectory theory also needs (Wiseman & Gambetta 2012), not a choice made for convenience. Given that specification, release is a property of the dynamics: a record is released when it commutes with the system’s accessible future through every return path the interaction network provides. Paper II bounds the phase any accessible observable can recover by a commutator. Where this is weakest is the apparatus-free case, Objection 6, and ACT lists it as its sharpest open item rather than as settled. One research lead is recorded: grounding accessibility in outgoing massless radiation, which would make the criterion objective at the cost of moving ACT toward ETH.

Carried by: ledger: the accessible algebra; the release bound; Evidence, where the event can sit; Mathematics, Part I.

“What happens when nothing was built to record? A dust grain in the microwave background, the early universe.”

Open, and the first item on ACT’s list. The selection theorem is proved with a record architecture in place: in a model where records form faster than the probe disperses, the bath’s fragments support only localization-type conditioning. What fixes the accessible algebra for a dust grain decohering in the thermal photon bath, where nothing was engineered, is not yet derived. The conjecture ACT states is alignment: that the alternatives the records select coincide with the pointer states the dynamics make robust. It is exact in the nondemolition class, coherent-state pointers in quantum Brownian motion are the candidate answer, and a regime with small nondemolition-breaking and large misalignment would refute it.

It is also the open item an experiment can reach. A levitated particle in a natural thermal bath, with the recorded observable reconstructed from environment fragments as in the quantum-Darwinism experiments, asks exactly what the environment selects when nothing is built to ask.

Carried by: ledger: selection without an apparatus; the alignment conjecture; Evidence, experiments worth doing.

“Whether an event has happened depends on the future of the interaction. Isn’t that retrocausal?”

It depends on the future dynamics, not on future outcomes, and that is the difference. In Paper II’s delayed-return model an atom’s emission comes back from a mirror: the in-flight record is nondemolition now but not released, and anchoring waits for the exit record. The laboratory shows the same structure: light returned by a mirror enhances or inhibits an ion’s own spontaneous emission (Eschner et al. 2001), and a qubit whose phonons return decays non-exponentially (Andersson et al. 2019). Which couplings exist, and so which return paths, is a fact about the apparatus at the time, not about anything that has yet to happen.

Nothing later alters an anchored record, and no information travels backward. The delayed-choice eraser is the test case. The later record on the idler side never changes the earlier detection of its partner; it decides which subset that detection belongs to. The unsorted signal distribution carries no interference and permits no signalling. ACT shares this trait with Kent’s late-time records and ETH’s algebra of the future; it is a family feature of event ontologies, stated here openly.

Carried by: ledger: delayed return; release; The Picture, the same resolution as delayed choice; Evidence, where the event can sit.

“The hazard is defined only for Markovian baths. What about a structured environment that hands phase back late?”

Paper I defines irreversibility as recoverable phase distinguishability over the accessible algebra, $\Phi_{\rm irr}$, and proves it nondecreasing when the accessible dynamics is CP-divisible; there the hazard $h=\dot\Phi_{\rm irr}$ is defined, and outside that condition it is not. A bath with a gap leaves $\Phi_{\rm irr}$ bounded; a single mode gives back what it took; the quantum step itself fails without the bath’s noise. These are the regimes in which no event law should fire, and ACT’s does not.

Beyond the Markov class the general event law of Paper II is the instrument path measure, and the one non-Markovian case worked through is delayed return, where anchoring is deferred to the exit record. Its continuum limit is open, because the delayed quantum stochastic equation is not of Hudson–Parthasarathy form, and the ledger lists non-Markovian release as open with it.

Carried by: ledger: the CP-divisible scope; $\Phi_{\rm irr}$; non-Markovian release; Mathematics, open problems.

“It makes no new prediction. Why believe it?”

Its predictions are those of quantum mechanics, by construction: the dynamics are untouched, and the event sits at release, where it changes no prediction. No anomaly is relied on, and no discriminating experiment is claimed. Every detection is one mark, as every account that keeps quantum mechanics unmodified predicts, and ACT takes that mark at face value.

What the experiments did instead is mapped on the Evidence page. They eliminated the neighbouring rules: an event at first interaction (Brune et al. 1996), at decoherence of the reduced state (Chapman et al. 1995; Kim et al. 2000), or at the first appearance of a record carrier (Eschner et al. 2001; Andersson et al. 2019). That leaves release as the earliest placement the data allow. They measured the mechanism: the event rate is read off the same coupling that draws the decoherence curve, with no free constant, and in the nondemolition case the two curves are one. And they could have refuted ACT along with unmodified quantum mechanics, had the excess radiation the collapse models predict appeared; it did not. Corroborated, not singled out, is the exact standing, and three experiments could still embarrass it: the Rock’s last slab, mapping the gate, and the particle in a natural bath.

The case for ACT is then explanatory. It says where definiteness lives, in released records, when Copenhagen, QBism and relational quantum mechanics decline to say, consistent histories make it framework-relative, and Everett denies there is anything to locate. It drops the particle, and the famous mysteries with it. It makes probability ordinary. And it lists what it adds.

Carried by: ledger: corroborated, not singled out; Evidence, which event rules survive; the collapse searches; experiments worth doing.

“Is it local? What does it say about Wigner’s friend? And is release covariant?”

It is not local, and it does not claim to be. One actual history with Born weights violates Bell inequalities in Bell-test geometries, so without superdeterminism or retrocausation ACT is nonlocal in Bell’s sense. It is nonlocal in the mildest form a one-history theory can take: outcome independence fails, parameter independence holds, and the correlation carries no signal. Parameter independence is not read off the weights: it is a requirement the theory places on the chance of a record, for every division of the world into here and elsewhere, and it is one of the two requirements that force the Born weights (Objection 4). Atemporality removes the time order of distant records, not the nonlocality.

A friend whose laboratory is kept coherently reversible has no anchored outcome during that interval: no record is released, so no event is assigned. The extended Wigner’s-friend inequalities, violated so far with photons and with GHZ states of up to 17 qubits, are consistent with that. In the Local Friendliness classification ACT gives up Friendliness: it keeps universal quantum theory, so it predicts that the violation will be observed with such a friend, and it denies that a cognition run coherently on a quantum computer has an observed event during the coherent interval, because no record is released. Any account that denies Friendliness for such a friend shares that cell; what is ACT’s own is the reason, release, and its application to a human friend whose laboratory is kept coherent. It is a definite commitment, stated here so that no one has to discover it.

Covariant release is open. Release is defined against a time-slice’s accessible future, and “central relative to the accessible future” has no covariant statement yet. Fewster–Verch instruments give a partially ordered record history, and Kent’s 2014 construction shows that a one-history ontology over unmodified unitary dynamics can be Lorentz covariant. ACT’s open problem is a covariant criterion, not covariant statistics, which it has.

Carried by: The Picture, entangled records and Bell; ledger: Bell; Wigner’s friend; covariant release.

“Isn’t Many Worlds simpler? It adds nothing to the equation.”

Every account in the count on The Picture but the collapse models reproduces the statistics of quantum mechanics exactly; the collapse models depart from them, and survive only where the radiation searches have left their constants open. Where the experiments cannot choose, Occam’s razor decides. The razor counts what must be assumed to account for what is observed, and what is observed is one outcome in every run, with Born frequencies.

Many Worlds adds nothing to the equation. It reads the equation as every outcome happening: a world for every outcome of every quantum event, with a copy of you in each, and no experiment in one world can reach another. It accounts for the one outcome you see by those copies, and it has to rebuild what a chance is; its derivations of the Born rule are still contested.

ACT adds one postulate, that one history of marks is actual, so the outcome a run shows is the only one that occurs; where exclusivity is not yet a theorem, it is stated as a second. Every account fixes its weights by an assumption of its own, Bohm’s equilibrium hypothesis and the Everettian decision axioms among them; ACT’s, given refinable records, are two requirements on its law of chance, no signalling and continuity. Its open problems are parts of the law still to be written, not added entities.

Both keep the whole wave, and an Everettian answers that the worlds are only its decohered components, which ACT keeps too. The difference is what a component holds. On ACT it holds possibilities for events and no events, and no copy of you who saw another outcome; on Many Worlds it holds one. No run can reach another component to check, and where no run can, the razor counts against positing copies of you that no experiment can find. Among the accounts in which an outcome is a fact apart from any agent or system, the count comes down to one comparison: ACT asserts one outcome for each run, the one the run shows; Many Worlds asserts that outcome and every other besides, the others in worlds no run can reach. The count for every account is on The Picture and in the flagship’s §9.

Carried by: The Picture, what each account adds; Objection 2; ledger: one history actual; exclusivity, a theorem in class; Born weights forced.

1. K. Sonderegger, From Matter Waves to Anchored Events (Paper I, 2026), §§6, 9–10, 12, 14–15; Released records (Paper II, 2026).
2. M. Brune et al., Phys. Rev. Lett. 76, 1800 (1996); M. S. Chapman et al., Phys. Rev. Lett. 75, 3783 (1995); Y.-H. Kim et al., Phys. Rev. Lett. 84, 1 (2000); J. Eschner et al., Nature 413, 495 (2001); G. Andersson et al., Nat. Phys. 15, 1123 (2019).
3. W.-K. Rhim, A. Pines & J. S. Waugh, Phys. Rev. B 3, 684 (1971); X. Mi et al., Science 374, 1479 (2021); H. M. Wiseman & J. M. Gambetta, Phys. Rev. Lett. 108, 220402 (2012).
4. H. R. Brown & D. Wallace, Found. Phys. 35, 517 (2005); D. Wallace, The Emergent Multiverse (Oxford University Press, 2012).
5. W. Heisenberg, Physics and Philosophy (Harper & Row, 1958); R. Haag, Commun. Math. Phys. 132, 245 (1990); V. P. Belavkin, Found. Phys. 24, 685 (1994); A. Kent, Phys. Rev. A 90, 012107 (2014); J. Fröhlich & A. Pizzo, Commun. Math. Phys. 389, 1673 (2022); J. Fröhlich, Z. Gang & A. Pizzo, Commun. Math. Phys. 406, 195 (2025); S. Zhang, arXiv:2607.13096 (2026).
6. J. P. Jarrett, Noûs 18, 569 (1984); K.-W. Bong et al., Nat. Phys. 16, 1199 (2020); Zeng, Labib & Russo, Quantum 9, 1851 (2025); Laux & Cavalcanti, arXiv:2609.12527 (2026, preprint); H. M. Wiseman, E. G. Cavalcanti & E. G. Rieffel, Quantum 7, 1112 (2023).
7. E. Aprile et al. (XENON), Phys. Rev. Lett. 136, 120201 (2026); S. Donadi et al., Nat. Phys. 17, 74 (2021).