Evidence
What the laboratory shows
Four thought experiments carry the argument, and each has been performed in the laboratory. In the Rock, a lossy slab that nothing reads still kills the fringes, so a bath, not an observer, keeps the record; in the Valves, the rate at which the fringes fade can be dialled; in the Filter, an apparatus picks components out of one wave, as a radio picks a station out of the air, and no event occurs until a record is released; in the Mark, single marks build a wave’s fringes when no record of the path was kept, and a particle pattern when one was. With the erasers, these experiments rule out each earlier place tested for an event of ACT’s kind; release is the earliest the data allow. What sets ACT apart from the other accounts that keep quantum mechanics unmodified is not a different prediction but the account it gives. Together the experiments show its physical ingredients: an interfering quantum field, a bath that carries records, and localized marks. ACT describes the electron as a stable wave-mode structure of the field, never a particle. Its modes evolve and interfere; the bath carries a record of the alternatives beyond coherent recovery; and the released record is an event: the mark. Under each picture, the experiments that bear on it can be opened row by row, each with what it corroborates and what it does not establish; the record in numbers — one hundred experiments, from 1922 to 2026 — follows the four.
Each thought experiment says what it defends in the Original Picture and what would have refuted it; each experiment under it says what was observed, what it corroborates, and what it does not establish. The last column is the one most often skipped, so it is kept on every row: it is what makes the first two worth believing.
The argument, in four pictures
The four are read in order, and together they walk the chain of the Picture: field, bath, record, rate, one history. The Rock finds the source of the decoherence: not an observer, but a bath. The Valves measure the bath: its rate can be dialled. The Filter shows the field as a composite: an apparatus picks out its components, and only a record picks the outcome. The Mark shows the event itself: every detection is one mark, and how the marks add up depends on whether a record was kept. After the four come the experiments that fix where an event can sit. Each exhibit gives the picture, what it shows, what could have refuted it, and the experiments, with what each corroborates and what it does not establish.
The Rock
Remove the observer. The record remains.
The picture
Send electrons through a double slit and they build fringes. Take away the detector at the slits and put a plain slab beside one path, with nothing reading it. The fringes die anyway. A resistive slab kills them. A good conductor barely touches them. An ideal superconductor should leave them alone. Nothing observed anything. The slab’s own electrons, stirred by the passing charge, carried a record of the path off as heat. That is the bath. For a slab, the record follows the dissipation: no dissipation, no record.
What it shows
It finds the source of the decoherence: not an observer, but the slab itself, a bath of electrons and vibrating atoms that takes up a trace of the path and carries it off as heat. That is principle III and sentence 3 of the Picture: the bath explains the detector effect, and a detector is any body that keeps an irreversible record of the path, whether or not anything reads it. It also gives the bath its first measured property: for a slab, the record follows the dissipation, strong in doped silicon, weak in gold, and expected to vanish in the superconducting limit. Interaction is not the criterion; a record is. This is the sense in which decoherence is built in: the same dissipation that keeps the record is the decoherence (Mechanism). The fringes themselves say that mass is present before anchoring: no cut set by mass or size has been found, and ACT, which has no isolated-system collapse, predicts none (The Picture). ACT reads each run as anchored at the slab or not yet anchored; the contrast that remains is the share of runs not yet anchored (the Mark).
Could have failed if
Interaction without a record had killed interference (it does not: gold, and atoms passing a surface, keep their fringes), or a dissipative body with no readout had left the fringes intact (it does not: silicon kills them), or a heavier fringe had failed to appear where quantum mechanics predicts it.
The experiments · 10 rows, each with what it corroborates and what it does not establish
| Experiment | Observed | Corroborates | Does not establish |
|---|---|---|---|
| Sonnentag & Hasselbach 2007 | Electrons in a biprism interferometer fly a few micrometres above a doped-silicon plate that nothing reads; contrast falls as the height shrinks and the path separation grows, to zero near the surface | A passive dissipative body keeps a which-path record | That an event occurred in any one run; the ensemble is a mixture |
| Kerker et al. 2020 | The same geometry with doped silicon and with gold, whose resistivity is about a million times lower; gold gives far weaker loss; macroscopic quantum electrodynamics fits without a fit factor and three competing models are excluded | Record strength tracks dissipation; a nearly lossless body keeps almost no record | The zero-loss (superconducting) limit, planned but not measured |
| Beierle, Zhang & Batelaan 2018 | Electron diffraction beside silicon and gold surfaces in a second apparatus; gold agrees with the same theory, and the silicon discrepancy is traced by Kerker to an approximation | Independent confirmation of the dissipation ordering | Final selection among the Coulomb-decoherence theories |
| Chapman et al. 1995 | One scattered photon destroys contrast; selecting the photon directions that leave the path undetermined restores it | The record, not the kick, is what matters | A disturbance-free reading, which is still debated |
| Dürr, Nonn & Rempe 1998 | Which-way information stored in internal states; the momentum back-action is too small to explain the loss | The same | The same |
| Hornberger et al. 2003 | C70 contrast falls with background-gas pressure as collisional theory predicts | A gas is a rock; the record is carried off in the gas | Anything beyond decoherence theory |
| Hackermüller et al. 2004 | Hot C70 loses coherence by emitting thermal photons | The molecule’s own radiation is a rock | The same |
| Liu et al. 2015 | Einstein’s recoiling slit realized in O2: the two atoms are the slits, and the ion’s own recoil keeps the which-slit record | The record can live in the slit itself | The same |
| Arndt 1999 · Gerlich 2011 · Fein 2019 · Pedalino 2026 | Matter-wave interference at ever greater mass: C60, molecules of 6,910 u, molecules beyond 25 kDa, and sodium nanoparticles of more than 7,000 atoms, mean mass 172 kDa, with fringes spaced by $\lambda=h/mv$ throughout | The wave is a wave at every mass yet measured; mass gives the field a clock before anything is anchored; each heavier fringe is a test ACT passes with quantum mechanics and against the collapse models | Nothing that separates ACT from quantum mechanics |
| Colella 1975 · Kovachy 2015 · Friedman 2000 · van der Wal 2000 · O’Connell 2010 · Delić 2020 · Bild 2023 | Neutrons feel gravity on their mass while in superposition; atomic wave packets separated by 54 cm still interfere; superconducting loops carry superposed opposite currents; a mechanical resonator is controlled phonon by phonon, a levitated nanoparticle is cooled to its motional ground state, and a 16 µg oscillator is put in a cat state | The same, in massive and macroscopic systems; no isolated-system collapse at any scale yet reached | The same |
Controls: the no-record limit · 2 rows
| Control | Observed | What it shows |
|---|---|---|
| Perreault & Cronin 2005 | Atoms through a nanograting pick up van der Waals phase shifts; interference intact | Interaction without a record leaves coherence alone |
| Kerker 2020, the gold run | The same Coulomb interaction with a million times less dissipation; almost no fringe loss | The record, not the interaction, sets the loss |
The Valves
Tune the bath. The rate follows.
The picture
Keep the rock and add a valve: the pressure of a gas, a laser that heats the molecules, the current in a detector beside one path, the gap between the paths. Turn it, and the fringes fade faster or slower, smoothly, by one rule in every kind of bath: faster for a stronger coupling, a hotter bath and, up to a point, a wider gap. The bath has a rate, and the laboratory can dial it. One valve is a fake: classical noise injected from outside dims the fringes in the average but records nothing, and ACT counts it as no event.
What it shows
It measures the bath. The rate at which a record forms is physical: it is set by the bath’s coupling and temperature and by the gap between the paths, and nothing in it is a free constant. Quantum Brownian motion, the bath’s exactly solvable model and the one Paper I builds its law on, writes the rule in one line for small separations at high temperature: the decoherence rate is the damping rate times half the square of the separation measured in thermal wavelengths, so, at a given gap and temperature, the decoherence cannot be dialled without dialling the damping (Mechanism). In Brune’s cavity the decoherence quickened as the separation grew, as the theory that ties it to the damping predicts. That is sentence 4 of the Picture: detector parameters set the anchoring rate, and the experiments already existed. The fake valve marks the line ACT draws: its event law counts only phase that can no longer be recovered, so classical noise that records nothing, and a dephasing an echo undoes, make no events; a quantum bath, whose record has left, does.
Could have failed if
Decoherence had been all-or-nothing, a fixed universal constant, or independent of the coupling; or if any experiment had produced irreversible loss with no record anywhere.
The experiments · 14 rows, each with what it corroborates and what it does not establish
| Experiment | Valve | Observed | Corroborates | Does not establish |
|---|---|---|---|---|
| Brune et al. 1996b | Path separation against the cavity’s leak | A cavity-field meter in a superposition of two phases decoheres progressively, faster for larger separation, as the theory in which one coupling gives both decoherence and damping predicts | The Caldeira–Leggett identity; the rate is physical | Anything beyond decoherence theory |
| Myatt et al. 2000 | Engineered reservoir strength | Ion motional superpositions decohere at rates dialled by engineered reservoirs; for the amplitude reservoir the rate grows as the separation squared | The rate law: the average fades at a rate set by the reservoir’s strength, growing as the separation squared; its laser-cooling reservoir is a real bath | A record from its random-field reservoirs: that noise is classical, the fake valve, and ACT counts no events (see the control) |
| Turchette et al. 2000 | Reservoir type | Amplitude, phase and zero-temperature (laser-cooling) reservoirs, each with its own rate law | The same; the laser-cooling reservoir is a real bath, its scattered photons carry the record | The same |
| Deléglise et al. 2008 | Time | A Wigner-function movie of a cat state decohering; the fringes decay in 17 ± 3 ms against a predicted 22 ms | Quantitative confirmation of the rate | The same |
| Pfau et al. 1994 | Zero to one photon | The spatial coherence of helium atoms is lost by a single spontaneous emission | The valve’s first notch is one record carrier | The same |
| Kokorowski et al. 2001 | Number of photons | Contrast against separation shows revivals near one scattered photon; they vanish as the number grows | Distinguishability, not interaction count | The same |
| Buks et al. 1998 | Detector current | Electron interference dephased continuously by the current through a quantum point contact beside one arm | A detector’s record strength is a dial | The same |
| Itano 1990 · Fischer 2001 · Signoles 2014 | Measurement frequency | Repeated measurement freezes a driven transition; the tunnelling of cold atoms is both suppressed (Zeno) and enhanced (anti-Zeno) by the interval between measurements; measurement confines a dynamics to a subspace | The valve changes the decay rate itself | The same |
| Patil, Chakram & Vengalattore 2015 | Continuous scattering rate | A lattice gas localizes as the imaging-light scattering rate rises | The same, continuously | The same |
| Jain et al. 2016 | Trap power | Photon-recoil heating of a levitated nanoparticle measured directly, scaling with laser power | A record channel with a measured rate in a massive object | The same |
| Gustafsson et al. 2014 | A phonon continuum | A superconducting qubit decays into surface-acoustic-wave phonons | The phonon valve, opened | The same |
| Andersson et al. 2019 | The bath’s memory | The same bath with a built-in delay; the decay goes non-exponential | The bath’s structure sets the rate | The same |
| Yin et al. 2013 | Open and close in real time | A resonator’s coupling to a transmission line switched in nanoseconds; a photon sent down the line is caught, held, and released on demand with a shaped waveform | The rate at which a record leaves is a valve, opened and shut in real time; what is caught is a photon from the line, not the resonator’s own emission returned | The same |
| Hornberger 2003 · Hackermüller 2004 (shared with the Rock) | Pressure; heating | The gas valve and the thermal-photon valve | Two calculable channels | The same |
Control: the virtual limit · 1 row
| Control | Observed | What it shows |
|---|---|---|
| Hahn 1950 | Spin echo: ensemble dephasing reversed by a π pulse | Classical dephasing records nothing and is undone; it is not record formation. ACT assigns it no events |
The Filter
The apparatus picks out components. Only a record picks the outcome.
The picture
An FM radio does not create the station it plays. The air carries one electromagnetic field, the sum of every broadcast at once. The tuner, a resonant circuit, picks out one frequency and leaves every other station on the air. The quantum field is a composite in the same way, a sum of modes, as a Fourier series is a sum of frequencies, and an apparatus can pick out its components.
A cavity is the tuner for an atom’s light. An atom can emit only into a mode the field offers. In a cavity tuned to its transition it emits faster; between plates that leave no mode at its frequency, its emission is suppressed. What leaves the atom is an excitation of a mode, not a projectile.
A Stern–Gerlach magnet sorts the same way. A beam of silver atoms splits in two along the magnet’s axis; so does a beam of hydrogen atoms, whose magnetism is the spin of a single electron. Turn the magnet and it sorts along the new axis, as a new setting of the dial brings in a new station. The magnet has found nothing out: in a full-loop Stern–Gerlach interferometer the two beams are brought back together, and they interfere. Which beam a detector finds the atom in is settled only when a record is released, with Born’s weights.
What it shows
So neither the atom nor its electron is a small body that takes one path or the other: each is a wave whose components can be separated and brought back together, and the particle is the mark left when a record of one component is released (the Mark). That is principle II: the field is a composite of modes, and what gets recorded is a mode. Which alternatives a record can register is set by the coupling, not chosen: a magnet sorts by spin along its axis, a cavity by frequency, and the environment copies records of only the alternatives its coupling selects across its fragments, the parts of it that can be read separately, as the redundancy experiments below show. Picking out components is unitary and reversible, so it is not the event: that is the event law’s what and which, the record algebra selected by the coupling before any event (Mechanism). The apparatus picks out components; only a record picks the outcome. Where the event can sit is set out after the Mark (below).
Could have failed if
The apparatus had fixed the outcome rather than the components; or a Stern–Gerlach separation had proved irreversible (it does not: the full loop recombines); or an atom had emitted the same way whatever modes its surroundings offer (it does not: a tuned cavity speeds the emission, and plates that leave no mode suppress it); or the pointer basis had been independent of the coupling (it is not: change the reservoir, change the basis).
The experiments · 14 rows, each with what it corroborates and what it does not establish
| Experiment | Observed | Corroborates | Does not establish |
|---|---|---|---|
| Gerlach & Stern 1922 | Silver atoms split into two beams along the gradient axis | The apparatus selects the axis; the outcome along it is random | Which alternative wins |
| Phipps & Taylor 1927 | The same with hydrogen: a single electron’s spin | The same, at the elementary level | The same |
| Frisch & Segrè 1933 | Sequential Stern–Gerlach stages with spin flips between them | A prepared mode answers the same question predictably | The same |
| Rabi, Zacharias, Millman & Kusch 1938 | Resonant transfer between spin states between a polarizer and an analyzer | Deterministic steering into a mode before the record | The same |
| Brossel, Kastler & Winter 1952 | Optical pumping drives atoms into one sublevel (proposed by Kastler in 1950) | The same, with light | The same |
| Goy, Raimond, Gross & Haroche 1983 | Spontaneous emission enhanced by a resonant cavity | The cavity selects which mode of the field an excitation may enter | The same |
| Hulet, Hilfer & Kleppner 1985 | Spontaneous emission inhibited when no mode is available | The same, in the other direction | The same |
| Gleyzes et al. 2007 | Dispersive, non-demolition coupling records photon number hundreds of times without absorbing the photon | The coupling, not the field, sets the record algebra | The same |
| Machluf, Japha & Folman 2013 | Coherent Stern–Gerlach momentum splitting on an atom chip | The splitting is a unitary, not an event | The same |
| Amit et al. 2019 | A Stern–Gerlach matter-wave interferometer whose phase grows as the cube of the time | The same | The same |
| Margalit et al. 2021 | A full-loop Stern–Gerlach interferometer: the spin-split beams recombine with high visibility | The magnet anchors nothing; the event is at the record | The same |
| Ciampini 2018 · Chen 2019 · Unden 2019 · Zhu 2025 | On photonic, nitrogen-vacancy and superconducting platforms, information about one pointer observable is encoded redundantly across many environment fragments, with the mutual information saturating the plateau quantum Darwinism predicts | What a record can register is an output of the coupling; in the fast-record regime, fragment records support only localization-type conditioning (Paper I’s selection theorem, Mathematics) | Which outcome occurs; quantum Darwinism explains agreement among observers, and the occurrence of one result is ACT’s primitive |
| Brune 1996b · Turchette 2000 (shared with the Valves) | Dissipative coupling makes coherent states the pointer states; a phase reservoir makes number states the pointer states, in the same ion | Change the coupling, change the pointer basis; the phase reservoir was classical noise, which shows which states survive and records nothing | The same |
| Chapman 1995 · Kokorowski 2001 (shared) | Coherence regained after a photon has flown; revivals after scatterings | The existence of a carrier is not the event; release is | The same |
The Mark
The wave is what exists. The particle is what happens.
The picture
Send electrons through two slits one at a time. Each arrives as a single mark, one dot on the screen. With no record of the path, the dots build up fringes: the pattern of a wave. Put the rock beside one path and the dots build up a plain band: the pattern particles would make. Every dot is the same kind of mark either way. The record changes how the marks add up because it changes when the first event of the run happens: at the rock, or at the screen. The wave is what exists. The particle is what happens.
What it shows
It is the headline, seen on a screen. Every detection is one mark, and no event leaves two. What a record of the path changes is how the marks add up: a wave’s fringes when no record was kept, the plain pattern particles would make when one was. So wave and particle are not two natures of one thing: the wave is what exists, always, and the particle is what happens, the mark left when a record is released (The Picture). It shows up in many ways: one photon at a beam splitter fires one detector, never both, though the same photons interfere; a single ion’s light switches on and off in steps, and after each click the clock restarts; the marks fall with Born’s weights; and the conditioned history of a single qubit has been reconstructed from its record. ACT reads this as the last link of the chain, anchoring to one causal history, and principle I: the mark is where a mode of the field enters causal time (Mechanism).
Two kinds of run. With no rock, nothing records the path before the screen. The dot is the first event of the run: the wave reaches the screen with both components, and the dot falls with the weight of their interference. With the rock, the path is recorded and released before the screen, since the rock’s heat is a record nothing can bring back. The run already has one event, the path, and the dot is its second: two events in one history, each leaving one mark. The dot follows the component that record selected; the other component no longer enters the run’s history, though the wave is untouched. A rock that records weakly anchors some runs and not others, and the screen shows the proportion: the fringes of the runs not yet anchored laid over the band of the runs anchored, with the contrast equal to the surviving fraction (Mechanism). So decoherence before the screen is anchoring before the screen whenever the record has been released. A record that is held but can still come back, a scattered photon’s for one, decoheres the reduced state just as fully and anchors nothing; the dot is then the first event of the run (where the event can sit).
Could have failed if
One event had ever left two marks; marks had been continuous rather than discrete; the marks had built fringes after a record of the path was released, or failed to build them when none was kept; the hazard had not restarted after a mark; or the frequencies had failed Born in any interference or non-demolition experiment.
The experiments · 11 rows, each with what it corroborates and what it does not establish
| Experiment | Observed | Corroborates | Does not establish |
|---|---|---|---|
| Merli, Missiroli & Pozzi 1976 | Single-electron buildup of biprism interference | One dot per electron; the pattern is the ensemble of dots | That these marks are the events and their history the one actual history: that is ACT’s one commitment, and no experiment decides it |
| Tonomura et al. 1989 | The same, electron by electron, filmed | The same | The same |
| Grangier, Roger & Aspect 1986 | A single photon on a beam splitter never fires both detectors; the same photons interfere in a Mach–Zehnder interferometer | One event, one mark, never both detectors; the wave goes both ways | The same |
| Kimble, Dagenais & Mandel 1977 | Antibunching in resonance fluorescence | After a mark the hazard restarts at zero. This did real work on ACT: Paper I’s hazard, extrapolated naively to a driven atom, predicts $g^{(2)}(0)=1$, and was replaced by the general event law of Paper II, which gives zero | The same |
| Diedrich & Walther 1987 | Antibunching from a single ion; $g^{(2)}(0)$ reaches zero | The same, one atom | The same |
| Nagourney, Sandberg & Dehmelt 1986 | Telegraph fluorescence of a single barium ion | Discrete, record-linked events | The same |
| Sauter, Neuhauser, Blatt & Toschek 1986 | The same, barium | The same | The same |
| Bergquist, Hulet, Itano & Wineland 1986 | The same, mercury; the dark-period statistics match the metastable lifetime | The same, with the right statistics | The same |
| Murch 2013 · Weber 2014 · Pokorny 2020 | Continuous weak measurement reconstructs the trajectory of a single superconducting qubit from its record; the most likely path between two states obeys a stochastic action principle; as the measurement strength is raised, coherence between measured and unmeasured levels disappears progressively while coherence inside the unmeasured subspace survives | The conditioned history ACT’s event law describes is an observable, manipulable object | That the trajectory is fundamental; it is reconstructed from the record, and that one such history is real is ACT’s commitment |
| Guerlin et al. 2007 | Progressive collapse of a cavity field to one photon number under repeated non-demolition probes; Born-distributed outcomes | One conditioned history, mark by mark, with Born weights | When, between probes, the event occurred: release-insensitive by theorem |
| Gleyzes 2007 (shared with the Filter) | A photon’s birth and death seen as jumps in hundreds of non-demolition readings | The record accumulates atom by atom; each detected atom is a released record | The same |
Where the event can sit
Earlier placements fail. Release survives.
The Filter’s magnet anchors nothing, and the laboratory has gone further. Here the experiments decide something outright: which event rules survive. Each rule below is read as an event rule of ACT’s kind: the coupled variable becomes actual at the moment named, and the actual history proceeds from that record, the unselected alternatives no longer entering it, while the global wave is untouched. A rule that assigns values at all times and lets the unselected alternatives keep entering the actual history, as Bohm’s does, is not of this kind and is not tested here. The laboratory has mapped the boundary directly, and ACT’s rule is the one left standing at the earliest place the data allow.
- An atom in a cavity takes its excitation back. A Rydberg atom coupled to a single cavity mode undergoes vacuum Rabi oscillation: it emits and reabsorbs, reversibly (Brune et al. 1996a).
- A record that can still be read back is not yet a fact. Which-path information held by a single photon can be erased, by choices made late or at spacelike separation, and the conditioned fringes return (Kim et al. 2000; Walborn et al. 2002; Ma et al. 2013, over 144 km). Interference reappears only after the signal detections are sorted by the corresponding idler outcomes; the unsorted signal distribution contains no interference and permits no signalling. A partial measurement of a superconducting qubit can be undone by a matched second measurement (Katz et al. 2006, 2008). Weak measurement followed by a reversal protects entanglement through amplitude damping (Kim et al. 2012). A quantum jump already under way can be caught and reversed in flight (Minev et al. 2019): the transition in flight is not yet an anchoring; the record clicks are.
- Emission is not final while the carrier can come back. Light returned by a mirror enhances or inhibits a trapped ion’s own spontaneous emission, depending on the mirror’s position (Eschner et al. 2001). A qubit coupled to surface acoustic waves at two points a hundred wavelengths apart decays non-exponentially, because the phonons return (Andersson et al. 2019).
- No record, no event until detection. In Wheeler’s delayed-choice experiments the decision to close the interferometer is made after a single quantum of light, or a single atom’s wave, has entered, and no environmental record is involved; the result follows the final configuration (Jacques et al. 2007; Manning et al. 2015). There was no particle path for the later choice to decide.
| When an event would happen | What that implies | Contradicted by | Status |
|---|---|---|---|
| At first interaction | An atom exchanging a photon with a cavity mode has “emitted or not” fixed at once; a magnet that splits a beam has decided the spin; a surface that shifts a phase has recorded a path | Brune et al. 1996a; Margalit et al. 2021; Perreault & Cronin 2005; the gold run of Kerker 2020 | Ruled out |
| When the reduced state decoheres | Which path is fixed, so interference could never be recovered by conditioning on the environment’s record | Chapman et al. 1995; Kim et al. 2000; Walborn et al. 2002; Katz et al. 2006, 2008; Minev et al. 2019 | Ruled out |
| When a record carrier exists | Emission is a finished fact, so a returning field could not alter the source’s own emission | Eschner et al. 2001; Andersson et al. 2019 | Ruled out |
| At release | The record is beyond coherent recovery by the system’s accessible future. Everything above, as observed | — | Survives (ACT) |
| Later, or never, or always | Same statistics, from interpretations without events, or with definite values at all times | — | Survive |
Two things follow. The data rule out every one-history rule of this kind that makes a record actual while the record can still coherently return, and ACT’s rule sits exactly on that boundary, at the earliest place the data allow. Each experiment in the table refuted a rule that placed the event at an independently identified earlier moment; ACT’s rule, which places it at the edge of recoverability, could only have failed had coherence returned from a record the account counts as released, and it has not. Rules placed later, and interpretations with no events at all, survive as well, because every interpretation that keeps the dynamics reproduces the same statistics. So the experiments fix ACT’s boundary and corroborate its placement; whether anything happens there is the one thing they leave to the postulate.
The reversibility boundary is also relative to what can be controlled. Spin dynamics spread through a solid have been time-reversed, and magnetization that seemed lost returned as an echo (Rhim, Pines & Waugh 1971; Pastawski et al. 2000); information scrambled across 53 superconducting qubits has been refocused (Mi et al. 2021), and across more since (Google Quantum AI 2025). ACT’s accessible algebra is therefore fixed by the physics of the coupling and the apparatus, not by convenience. Quantum-trajectory theory needs the same specification (Wiseman & Gambetta 2012).
One hundred experiments
Seventy-six published experiments corroborate the four thought experiments: nineteen the Rock, fifteen the Valves, thirty the Filter (fifteen on what an apparatus picks out, fifteen that map where the event can sit), twelve the Mark. Each is counted once even when it serves two exhibits. Two controls show the no-record limit: an interaction that keeps no record leaves coherence alone, and a dephasing an echo undoes was never a record. Eleven set the constraints any ontology must meet, and ACT meets them. Eleven searched for the heating and radiation that collapse models predict and found none — tests that ACT, which adds no dynamics, could have failed together with quantum mechanics, and did not. None of the hundred conflicts with the mechanism. Theory papers are cited for attribution and never counted.
| Thought experiment | Link in the chain | Establishes | Could have failed if | Experiments |
|---|---|---|---|---|
| The Rock | bath interaction → stable record | A bath, not an observer, keeps the record (ACT reads: principle III; sentence 3 of the Picture) | interaction without a record had killed fringes, or a lossy body with no readout had left them intact | 19, and 1 control |
| The Valves | the rate of record formation | The bath has a rate the laboratory can dial; the experiments already existed (ACT reads: sentence 4 of the Picture) | the rate had been fixed, all-or-nothing, or independent of the coupling | 15, 1 shared, and 1 control |
| The Filter | composite field → which mode can be recorded | One composite field: an apparatus picks out its components, and they can be brought back together while no record is released (ACT reads: principle II; the event law’s what and which; where the event can sit) | the apparatus had fixed the outcome, or a separation had been irreversible, or emission had ignored the modes on offer; or, where the event can sit, coherence had failed to return while a record could | 30, of which 15 map the boundary; 6 shared |
| The Mark | anchoring → one causal history | Every detection is one mark, and no event leaves two (ACT reads: the particle is what happens, the headline; principle I, the mark is where a mode enters time) | one event had ever left two marks, or the marks had built fringes after a record of the path was released, or the frequencies had failed Born | 12, and 1 shared |
What every interpretation must respect
- Bell tests. Loophole-free violations of Bell inequalities (Hensen et al. 2015; Giustina et al. 2015; Shalm et al. 2015; Storz et al. 2023, with $S=2.0747\pm0.0033$ across 30 m) rule out every theory in which each outcome depends only on its own detector and on something shared, with free settings. ACT is nonlocal in Bell’s sense, by outcome dependence only, with no signalling (The Picture).
- Leggett–Garg and contextuality. Leggett–Garg violations with ideal negative measurements (Palacios-Laloy et al. 2010; Robens et al. 2015) and state-independent contextuality (Kirchmair et al. 2009) rule out definite but unknown values. On ACT, what is not yet anchored has no value to be unknown.
- Extended Wigner’s friend. Inequalities that assume the absoluteness of observed events, locality and free settings are violated when the “friend” is a photon or a GHZ state of up to 17 qubits held reversibly (Proietti et al. 2019; Bong et al. 2020; Zeng, Labib & Russo 2025; Laux & Cavalcanti 2026). On ACT no record was released in those friends, so no event is assigned to them; a friend whose laboratory stays coherently reversible has no anchored outcome.
Tests ACT could have failed
Spontaneous-collapse models make risky predictions: a universal noise heats matter and makes it radiate. Underground germanium and xenon detectors see no excess (Donadi et al. 2021; Arnquist et al. 2022; Aprile et al. 2026), which excludes parameter-free Diósi–Penrose collapse and the original GRW rates under white-noise assumptions, and force-noise measurements in space (LISA Pathfinder: Carlesso et al. 2016; Helou et al. 2017; Altamura, Vinante & Carlesso 2025), on cantilevers (Vinante et al. 2016, 2017, 2020) and with levitated particles (Pontin et al. 2020; Zheng et al. 2020; Vinante et al. 2020; Di Bartolomeo & Carlesso 2024) close more of the parameter space. Coloured and dissipative versions remain alive (Carlesso, Ferialdi & Bassi 2018).
ACT adds no dynamics to quantum mechanics, so it predicts no excess radiation and no excess force noise; its noise and its damping are the bath’s own, tied by the fluctuation–dissipation relation (Mechanism). Had an excess been seen, ACT would have been refuted along with the unmodified dynamics it keeps. None was. This is ACT’s empirical vulnerability, the same as quantum mechanics’, and these experiments are the running test of it; so far ACT and the collapse models stand on opposite sides of the result.
Where ACT’s law can still be embarrassed
What can be tested is the event law in the regimes where it is stated regionally: not “is ACT true?” but “does the law extend, or break?” Three experiments do that work, and the running programme does the rest.
- Mapping the gate. Generalize Eschner et al. 2001 and Andersson et al. 2019: add a weak in-flight probe of the returning carrier with tunable readout strength, and measure how the return-induced modulation of the source’s emission degrades as the probe strengthens. It maps the release boundary continuously, in the configuration where Paper II says anchoring is deferred to the exit record and an in-flight click can be overturned; Paper II’s graded release bound is an inequality the data can check.
- Record selection without an apparatus. A levitated particle decohering in a natural thermal photon bath, with the redundantly recorded observable reconstructed from environment fragments as in the quantum-Darwinism experiments. It asks what the environment selects when nothing is engineered — the Claim Ledger’s sharpest open item — and the alignment conjecture, that record-selected alternatives coincide with the dynamically robust pointer states, is a claim such data can embarrass.
- Non-Markovian release. Giant atoms and structured reservoirs in circuit QED, where CP-divisibility fails and the event law currently says that no hazard is defined. Conditioned statistics there tell whether the law extends to a structured environment that returns a released record, or stops at its stated boundary.
- The Rock’s last slab. Kerker’s planned superconducting surface: the same electrons, the same geometry, no dissipation. ACT, with decoherence theory, says the fringes stay. A loss of contrast beside a lossless body would be a record with no carrier, and the Rock would fail.
- The running programme. Every heavier interference fringe and every underground radiation search is a test ACT could fail together with quantum mechanics. Each null result is ACT’s result; space-based matter-wave interferometry and larger xenon exposures are the continuation.
None of these distinguishes ACT from quantum mechanics. Each tests a claim ACT makes about its own event law in a regime the law has not yet covered, which is the only kind of test an interpretation that adds no dynamics can honestly propose.
What the laboratory has settled
ACT places an event only where a record has become irreversibly unavailable to the system’s accessible future. That boundary has been mapped independently: coherence returns exactly when, and only when, the record remains coherently recoverable. Event rules of ACT’s kind — rules under which the actual history proceeds from a record, the global wave untouched — that place the event at first interaction, at decoherence of the reduced state, or at the first appearance of a record carrier are already ruled out. No experiment distinguishes ACT from later event rules, or from interpretations that reproduce quantum mechanics without events.
What the laboratory has given ACT, in the four pictures above, is its mechanism measured, its boundary mapped, its placement tested by every recovery experiment and never refuted, and its one empirical vulnerability, the same as quantum mechanics’, still standing. Every detection in every experiment here is one mark, as every account that keeps quantum mechanics unmodified predicts, and ACT takes that mark at face value. ACT’s claims are about what exists and what happens, and the Claim Ledger says how each is held.
The four mysteries, beside the experiments
- Which slit did it go through? The Mark: with no record of the path, single marks build a wave’s fringes. The wave went through both slits; there was never a particle to go through one.
- Does the cat’s fate wait for someone to look? No. The Rock: a slab that nothing reads keeps the record. In any real box the Geiger counter’s bath makes the mark, whether or not anyone looks.
- Does a later choice reach back in time? Where the event can sit: in the erasers the later choice only sorts marks already made, and in Wheeler’s experiments no mark is made until the end. There was no particle path for it to decide.
- Are there copies of you in other worlds? The Mark: every detection is one mark, and no event has ever left two. ACT takes that mark at face value: one mark, one history, one of you.
The experiments fit this reading at every step. They do not rule out Bohm’s path or Many Worlds’ copies, which predict the same marks.
Grouped by exhibit. Theory papers, marked as such, are cited for attribution and are not counted.
- The Rock. P. Sonnentag & F. Hasselbach, Phys. Rev. Lett. 98, 200402 (2007); N. Kerker, R. Röpke, L. M. Steinert, A. Pooch & A. Stibor, New J. Phys. 22, 063039 (2020); P. J. Beierle, L. Zhang & H. Batelaan, New J. Phys. 20, 113030 (2018); M. S. Chapman et al., Phys. Rev. Lett. 75, 3783 (1995); S. Dürr, T. Nonn & G. Rempe, Nature 395, 33 (1998); K. Hornberger et al., Phys. Rev. Lett. 90, 160401 (2003); L. Hackermüller et al., Nature 427, 711 (2004); X.-J. Liu et al., Nat. Photon. 9, 120 (2015). M. Arndt et al., Nature 401, 680 (1999); S. Gerlich et al., Nat. Commun. 2, 263 (2011); Y. Y. Fein et al., Nat. Phys. 15, 1242 (2019); S. Pedalino et al., Nature 649, 866 (2026); R. Colella, A. W. Overhauser & S. A. Werner, Phys. Rev. Lett. 34, 1472 (1975); T. Kovachy et al., Nature 528, 530 (2015); J. R. Friedman et al., Nature 406, 43 (2000); C. H. van der Wal et al., Science 290, 773 (2000); A. D. O’Connell et al., Nature 464, 697 (2010); U. Delić et al., Science 367, 892 (2020); M. Bild et al., Science 380, 274 (2023). Control: J. D. Perreault & A. D. Cronin, Phys. Rev. Lett. 95, 133201 (2005). Theory: S. Scheel & S. Y. Buhmann, Phys. Rev. A 85, 030101(R) (2012); J. R. Anglin, J. P. Paz & W. H. Zurek, Phys. Rev. A 55, 4041 (1997).
- The Valves. M. Brune et al., Phys. Rev. Lett. 77, 4887 (1996) [1996b]; C. J. Myatt et al., Nature 403, 269 (2000); Q. A. Turchette et al., Phys. Rev. A 62, 053807 (2000); S. Deléglise et al., Nature 455, 510 (2008); T. Pfau, S. Spälter, Ch. Kurtsiefer, C. R. Ekstrom & J. Mlynek, Phys. Rev. Lett. 73, 1223 (1994); D. A. Kokorowski et al., Phys. Rev. Lett. 86, 2191 (2001); E. Buks, R. Schuster, M. Heiblum, D. Mahalu & V. Umansky, Nature 391, 871 (1998); W. M. Itano et al., Phys. Rev. A 41, 2295 (1990); M. C. Fischer, B. Gutiérrez-Medina & M. G. Raizen, Phys. Rev. Lett. 87, 040402 (2001); A. Signoles et al., Nat. Phys. 10, 715 (2014); Y. S. Patil, S. Chakram & M. Vengalattore, Phys. Rev. Lett. 115, 140402 (2015); V. Jain et al., Phys. Rev. Lett. 116, 243601 (2016); M. V. Gustafsson et al., Science 346, 207 (2014); G. Andersson et al., Nat. Phys. 15, 1123 (2019); Y. Yin et al., Phys. Rev. Lett. 110, 107001 (2013). Control: E. L. Hahn, Phys. Rev. 80, 580 (1950). Theory: A. O. Caldeira & A. J. Leggett, Physica A 121, 587 (1983).
- The Filter. W. Gerlach & O. Stern, Z. Phys. 9, 349 (1922); T. E. Phipps & J. B. Taylor, Phys. Rev. 29, 309 (1927); R. Frisch & E. Segrè, Z. Phys. 80, 610 (1933); I. I. Rabi, J. R. Zacharias, S. Millman & P. Kusch, Phys. Rev. 53, 318 (1938); J. Brossel, A. Kastler & J. Winter, J. Phys. Radium 13, 668 (1952), and A. Kastler, J. Phys. Radium 11, 255 (1950); P. Goy, J. M. Raimond, M. Gross & S. Haroche, Phys. Rev. Lett. 50, 1903 (1983); R. G. Hulet, E. S. Hilfer & D. Kleppner, Phys. Rev. Lett. 55, 2137 (1985); S. Gleyzes et al., Nature 446, 297 (2007); S. Machluf, Y. Japha & R. Folman, Nat. Commun. 4, 2424 (2013); Y. Amit et al., Phys. Rev. Lett. 123, 083601 (2019); Y. Margalit et al., Sci. Adv. 7, eabg2879 (2021); M. A. Ciampini et al., Phys. Rev. A 98, 020101(R) (2018); M.-C. Chen et al., Sci. Bull. 64, 580 (2019); T. K. Unden et al., Phys. Rev. Lett. 123, 140402 (2019); Z. Zhu et al., Sci. Adv. 11, eadx6857 (2025). Theory: E. M. Purcell, Phys. Rev. 69, 681 (1946).
- The boundary. M. Brune et al., Phys. Rev. Lett. 76, 1800 (1996) [1996a]; Y.-H. Kim et al., Phys. Rev. Lett. 84, 1 (2000); S. P. Walborn et al., Phys. Rev. A 65, 033818 (2002); X.-S. Ma et al., Proc. Natl. Acad. Sci. USA 110, 1221 (2013); N. Katz et al., Science 312, 1498 (2006); Phys. Rev. Lett. 101, 200401 (2008); Y.-S. Kim et al., Nat. Phys. 8, 117 (2012); Z. K. Minev et al., Nature 570, 200 (2019); J. Eschner et al., Nature 413, 495 (2001); V. Jacques et al., Science 315, 966 (2007); A. G. Manning et al., Nat. Phys. 11, 539 (2015); W.-K. Rhim, A. Pines & J. S. Waugh, Phys. Rev. B 3, 684 (1971); H. M. Pastawski et al., Physica A 283, 166 (2000); X. Mi et al., Science 374, 1479 (2021); Google Quantum AI and Collaborators, Nature 646, 825 (2025). Theory: H. M. Wiseman & J. M. Gambetta, Phys. Rev. Lett. 108, 220402 (2012).
- The Mark. P. G. Merli, G. F. Missiroli & G. Pozzi, Am. J. Phys. 44, 306 (1976); A. Tonomura, J. Endo, T. Matsuda, T. Kawasaki & H. Ezawa, Am. J. Phys. 57, 117 (1989); P. Grangier, G. Roger & A. Aspect, Europhys. Lett. 1, 173 (1986); H. J. Kimble, M. Dagenais & L. Mandel, Phys. Rev. Lett. 39, 691 (1977); F. Diedrich & H. Walther, Phys. Rev. Lett. 58, 203 (1987); W. Nagourney, J. Sandberg & H. Dehmelt, Phys. Rev. Lett. 56, 2797 (1986); Th. Sauter, W. Neuhauser, R. Blatt & P. E. Toschek, Phys. Rev. Lett. 57, 1696 (1986); J. C. Bergquist, R. G. Hulet, W. M. Itano & D. J. Wineland, Phys. Rev. Lett. 57, 1699 (1986); K. W. Murch et al., Nature 502, 211 (2013); S. J. Weber et al., Nature 511, 570 (2014); F. Pokorny et al., Phys. Rev. Lett. 124, 080401 (2020); C. Guerlin et al., Nature 448, 889 (2007).
- Constraints. B. Hensen et al., Nature 526, 682 (2015); M. Giustina et al., Phys. Rev. Lett. 115, 250401 (2015); L. K. Shalm et al., Phys. Rev. Lett. 115, 250402 (2015); S. Storz et al., Nature 617, 265 (2023); A. Palacios-Laloy et al., Nat. Phys. 6, 442 (2010); C. Robens et al., Phys. Rev. X 5, 011003 (2015); G. Kirchmair et al., Nature 460, 494 (2009); M. Proietti et al., Sci. Adv. 5, eaaw9832 (2019); K.-W. Bong et al., Nat. Phys. 16, 1199 (2020); Zeng, Labib & Russo, Quantum 9, 1851 (2025); Laux & Cavalcanti, arXiv:2609.12527 (2026, preprint).
- Collapse models. S. Donadi et al., Nat. Phys. 17, 74 (2021); I. J. Arnquist et al., Phys. Rev. Lett. 129, 080401 (2022); E. Aprile et al., Phys. Rev. Lett. 136, 120201 (2026); M. Carlesso et al., Phys. Rev. D 94, 124036 (2016); B. Helou et al., Phys. Rev. D 95, 084054 (2017); Altamura, Vinante & Carlesso, Phys. Rev. A 111, L020203 (2025); A. Vinante et al., Phys. Rev. Lett. 116, 090402 (2016); 119, 110401 (2017); 125, 100404 (2020); A. Pontin et al., Phys. Rev. Research 2, 023349 (2020); D. Zheng et al., Phys. Rev. Research 2, 013057 (2020); A. Vinante et al., Phys. Rev. Research 2, 043229 (2020); G. Di Bartolomeo & M. Carlesso, New J. Phys. 26, 043006 (2024). Theory: M. Carlesso, L. Ferialdi & A. Bassi, Eur. Phys. J. D 72, 159 (2018).