Is the Wavefunction a Real Physical Object or Just Our Knowledge?

Imagine you are handed a sealed envelope and told it contains either a photograph of a cat or a blank sheet of paper. You cannot open it, but you can perform certain restricted tests on it that partially reveal, without fully disclosing, what is inside. Now imagine that no matter how cleverly you devise these tests, no matter how many envelopes you compare, you can never determine whether the envelope itself is the picture — a physical thing with the cat's image etched into its fibers — or whether the envelope merely encodes your own uncertainty about a fact that already has a definite answer somewhere else. This is, in miniature, the oldest standing argument in the house of quantum mechanics. The envelope is the wavefunction. Physicists have spent a century arguing about what is actually inside it, and the argument is nowhere near over.

The peculiar thing is that this is not a fringe dispute. It is a disagreement among people who agree completely on the mathematics, on every predicted number, on every experimental outcome to twelve decimal places. Two physicists can run the identical calculation, get the identical answer, publish in the same journal — and hold irreconcilable views about whether the central object in the calculation, ψ, refers to something that exists in the world or to something that exists in their heads. That kind of disagreement, persisting undiminished for a hundred years despite total empirical agreement, is rare in the history of science. It is also, arguably, the deepest fissure running under all of quantum theory.

What makes the question so stubborn is that it appears, at first, almost unanswerable by experiment. If two positions make exactly the same predictions for every measurement anyone can perform, in what sense could evidence ever decide between them? For much of the twentieth century, this looked like a question for philosophers, not physicists — a matter of interpretation, safely quarantined from the equations. That quarantine has broken down. Since the early 2000s, a sequence of precise, mathematically rigorous no-go theorems has shown that the interpretations are not, after all, empirically equivalent in every conceivable sense. The walls of the envelope have started to reveal a little about what they can and cannot contain.

svg[Two readings of the same quantum recipe]

:::The same formalism, two metaphysical pictures. The PBR theorem attacks the right-hand picture directly.

For decades this dispute lived comfortably in the seminar room, seemingly beyond experimental reach, because both pictures could be dressed up to reproduce every known prediction. Then, in 2012, Matthew Pusey, Jonathan Barrett, and Terry Rudolph published a short paper that changed the terms of the debate. Their result, now called the PBR theorem, does something remarkable: it takes seriously the idea that ψ might be purely epistemic — that distinct quantum states could correspond, at least sometimes, to overlapping underlying physical states λ — and shows that this assumption, combined only with two extremely modest hypotheses, leads to predictions that contradict quantum mechanics.

The two hypotheses are worth stating plainly, because the whole force of the theorem rests on how mild they seem. First, there exists some underlying physical state λ, and the quantum state ψ is a (possibly incomplete) description of it — a "psi-epistemic" hidden-variable model in the broad sense. Second — and this is the crucial technical assumption — the physical states λ compatible with different, independently prepared quantum states are statistically independent of each other; preparing two separate systems in states ψ₁ and ψ₂ does not conspire to correlate their underlying λ's in some hidden way. Given just this, PBR show that if two distinct quantum states ψ₁ and ψ₂ could ever correspond to overlapping distributions over λ — if the "envelopes" could sometimes contain the same underlying picture — then a suitably designed joint measurement on several copies of the systems would produce outcomes that quantum mechanics forbids with certainty. Since we observe exactly the outcomes quantum mechanics predicts, the overlap must not occur: for any two distinct pure states, their λ-distributions must be entirely disjoint.

That conclusion sounds abstract, but its consequence is sharp. If the λ-distributions never overlap, then knowing λ is equivalent to knowing ψ — you could, in principle, read the exact quantum state directly off the underlying physical state. In other words, ψ cannot be a matter of mere partial ignorance layered over a richer reality that erases the distinction between quantum states; the quantum state must, in this framework, be recoverable from — essentially identical to — whatever the "real" physical state is. This is a formal argument for a strongly ontic reading of ψ, at least within any theory satisfying PBR's two assumptions.

It is important to be honest about the theorem's limits, because they are exactly where the live controversy now sits. PBR's independence assumption — often called "preparation independence" — is not logically forced on us; it is a further physical hypothesis, and it can fail. Theories in which the hidden states of independently prepared systems are secretly correlated, a property sometimes discussed under the banner of superdeterminism or retrocausal influence, can dodge PBR's conclusion entirely, in the same family of ways that superdeterminism can dodge Bell's theorem. Most physicists find such theories deeply unattractive because they seem to sacrifice the ordinary notion of free experimental choice or demand conspiratorial fine-tuning between the state of the universe at its origin and an experimenter's decisions billions of years later — but "unattractive" is not "impossible," and the loophole is logically real.

There is a second limitation. PBR's argument, in its cleanest form, concerns systems prepared independently; it says nothing directly about whether a single, already-entangled composite system's reduced state must be ontic in the same strong sense. And there are related but weaker no-go results — the earlier work of Lucien Hardy on a would-be proof, subtler theorems by Rob Spekkens and others exploring exactly how much "epistemicity" survives — that carve out a landscape rather than a single verdict. Spekkens, in particular, built explicit toy models in which many signature quantum phenomena, including some forms of interference and even weak analogues of Bell-type correlations, emerge naturally from an epistemic restriction: a hypothetical fundamental limit on how much any observer can ever know about an underlying classical ontic state, even in principle. Spekkens' toy theory does not reproduce all of quantum mechanics — critically, it fails to violate Bell inequalities the way real quantum systems do — but it reproduces a startling amount, enough to show that many of the "obviously quantum, must be ontic" phenomena that once seemed like knock-down arguments for realism are not, by themselves, decisive at all. This is precisely why the field could not simply declare the case closed after 2012: PBR closes one door quite firmly, while work like Spekkens' keeps a neighboring window open.

Meanwhile, the deeper conceptual sickness that makes any of this urgent — the measurement problem — has never gone away. Unitary Schrödinger evolution, left to itself, spreads superpositions outward without limit, entangling measuring devices, laboratories, and (in thought experiments) cats, observers, and the rest of the universe into ever-larger superpositions. Nothing in the bare formalism explains why we experience one single, definite outcome rather than a superposition of all of them, nor when the vaguely defined "collapse" is supposed to occur. Whether ψ is ontic or epistemic bears directly on how one even frames this problem: if ψ is a real physical state, collapse must be a real physical process demanding its own dynamical explanation (as attempted in objective-collapse theories); if ψ is a state of knowledge, "collapse" is nothing more mysterious than updating a probability distribution on new data, and the burden shifts instead to explaining why measurement outcomes are objectively definite at all, in a world with no wavefunction-as-thing to physically collapse.

The landscape of positions is more textured than a simple binary, and it helps to walk the main camps as they actually stand today, rather than caricature two teams.

Strict ontic realism treats ψ, quite literally, as a wave or field-like entity, generally understood as living not in ordinary three-dimensional space but in the vastly higher-dimensional configuration space of all the particles in a system. The Many-Worlds Interpretation (Everett's relative-state formulation, developed further by Bryce DeWitt, David Deutsch, and others) takes this to its most uncompromising conclusion: the universal wavefunction never collapses at all, evolves under the Schrödinger equation alone forever, and each branch of a superposition simply becomes, from the inside, an equally real, causally isolated world. Its appeal is mathematical austerity — no extra collapse postulate, no extra hidden variables, just the bare unitary theory taken at face value. Its trouble is notorious: explaining why the Born rule's probabilities should mean anything at all when every outcome literally happens somewhere, a problem that has generated an entire subliterature (decision-theoretic derivations by Deutsch and Wallace, "self-locating uncertainty" arguments) without full consensus that the puzzle is solved.

Objective collapse theories, such as the Ghirardi–Rimini–Weber (GRW) model and Roger Penrose's gravitationally-induced collapse proposal, also treat ψ as fully ontic — but they modify the Schrödinger equation itself, adding a small, stochastic, nonlinear term that causes spontaneous, random localization events. These events are exceedingly rare for isolated particles but become overwhelmingly frequent for macroscopic superpositions of many entangled particles, explaining why cats and pointers are never seen in superposition while single electrons plainly are. This is perhaps the most experimentally exposed position on the table: because it modifies the dynamics, it makes genuinely new predictions — tiny amounts of spontaneous heating, minuscule position diffusion, that in principle differ from standard quantum mechanics and are being hunted for in ultracold mechanical resonators and precision interferometry. Its cost is a degree of ad hoc-ness: the collapse rate and localization scale are free parameters tuned to match observation rather than derived from deeper principles, though Penrose's gravitational version at least motivates the scale from Planck-mass physics.

De Broglie–Bohm pilot-wave theory keeps ψ as a real, ontic guiding field but restores definite particle trajectories underneath it — an explicit hidden-variable theory in the PBR sense, and indeed a fully worked-out existence proof that empirically adequate hidden-variable completions of quantum mechanics are possible. Its appeal is conceptual clarity: there is no measurement problem in the usual sense, because particles always have definite positions, and "collapse" is just an effective description of what happens when the wavefunction's branches decohere and stop influencing the particle's actual trajectory. Its cost is a strange asymmetry between the wave (governing everything, itself an ontic entity in a high-dimensional configuration space with no matching field in ordinary space) and the particles (merely along for the ride), plus a famously awkward relationship with relativity, since the theory's dynamics rely on an explicit, preferred simultaneity structure that is hard to reconcile cleanly with Lorentz invariance.

QBism (quantum Bayesianism), developed principally by Christopher Fuchs, N. David Mermin, and Rüdiger Schack, occupies the most thoroughly epistemic pole. On this view ψ is not a state of the world at all, not even indirectly, but a personal, subjective degree of belief — a strictly Bayesian credence — that an individual agent uses to organize their own expectations for the outcomes of future measurements, including measurements the agent performs on their own body and instruments. Collapse, on this reading, is simply an agent updating their own beliefs the instant they get new data, no more mysterious or in need of physical mechanism than a poker player updating the odds after seeing a new card. QBism's appeal is that it dissolves the measurement problem and the collapse puzzle almost by definitional fiat, and it takes seriously — arguably more seriously than any rival — the fact that quantum mechanics is, operationally, a theory about an agent's expectations for experimental outcomes. Its trouble is that critics find it too deflationary: if ψ is purely personal, radically agent-relative belief, it becomes unclear why different agents' subjective wavefunctions should agree so exquisitely with each other and with the objective, mind-independent statistical regularities of nature — QBism's defenders answer this with an elaborate machinery built around the Born rule reformulated via symmetric informationally complete measurements (SIC-POVMs), arguing that agreement emerges from a shared underlying probabilistic structure of nature itself, but not everyone finds this fully satisfying.

Psi-epistemic hidden-variable models in the tradition PBR set out to constrain — of which Spekkens' toy models are the clearest existence proofs — occupy a middle ground: there is a real, objective, mind-independent λ underneath, and ψ is knowledge about λ in a fully realist, non-subjective sense (unlike QBism), but PBR shows any such model, if it also satisfies preparation independence, must make the λ-distributions for distinct pure states essentially non-overlapping — which arguably empties the "epistemic" label of much of its original content. Whether anything worth calling "epistemic" survives this constraint, or whether PBR has effectively pushed every serious hidden-variable model into a de facto ontic corner (differing from strict ontic realism only in retaining an underdetermined λ beneath ψ, as in Bohmian mechanics), is itself a live question inside the foundations community.

PBR's argument is a reductio: assume epistemicity plus independence, derive an impossible prediction, conclude overlap cannot occur — unless independence itself is dropped.

Because PBR's theorem is a mathematical proof, not an experimental result, one honest answer is: nothing new needs to be "discovered" for its logical content — it is already as settled as a proof gets, conditional on its assumptions. What remains open, and could in principle be constrained further, splits into several separate empirical and theoretical fronts.

The first front is closing the preparation-independence loophole in something like the way Bell tests have progressively closed locality and detection loopholes. This is extraordinarily hard, because preparation independence is not a straightforwardly testable statistical assumption in the way locality is; it is closer to a background metaphysical commitment about how independent preparation devices behave. Some researchers have explored weakened versions of PBR that assume only approximate, rather than perfect, independence, and shown the conclusion survives in a graded form — distinct quantum states can overlap only a little, with the allowed overlap shrinking as the independence assumption is tightened. Sharper experimental tests of these graded PBR-type inequalities, using multi-photon or trapped-ion platforms capable of extremely well-controlled independent state preparation, are an active and realistic near-term direction, plausibly advancing significantly within the coming decade.

The second front is testing the objective-collapse theories on their distinctive, falsifiable predictions. GRW-type models predict a faint but nonzero spontaneous heating and diffusion of massive objects, growing with the number of entangled constituent particles. Experiments levitating and cooling nanoparticles and micromechanical membranes to their motional ground state, or performing matter-wave interferometry with ever-larger molecules (the group of Markus Arndt has pushed such interferometry to molecules of tens of thousands of atomic mass units), are directly probing the regime where these models predict deviations from standard quantum mechanics should start to appear. So far, standard quantum mechanics has won every round, pushing the allowed collapse rate parameters into an ever-narrower band — but the parameter space is not yet closed, and proposed space-based interferometry missions (long discussed, including concepts building on the MAQRO proposal) could extend the reach by orders of magnitude, probing collapse models — including Penrose's gravitationally motivated version — much more decisively, plausibly within a few decades if such missions fly.

The third front is more conceptual than experimental: extending no-go results to cover cases PBR does not, especially entangled subsystems and cosmological or "one universe, no repeated preparations" contexts where the very notion of independent preparation loses its footing. If the wavefunction in question is the wavefunction of the entire universe — as it must be in a fully quantum cosmology — the PBR framework, built around comparing many independently prepared copies of a system, arguably cannot even be phrased. Making sense of ontic-versus-epistemic questions for a single, non-repeatable cosmic wavefunction, without begging the question either way, remains genuinely unresolved and connects directly to unsettled problems in quantum cosmology and quantum gravity.

A fourth, more radical, front concerns the relationship between this question and gravity itself. If ψ is fully ontic, then a superposition of two different mass distributions should, in principle, source two different gravitational fields in superposition — a strange but seemingly required consequence if gravity is to remain coupled to genuinely quantum matter in the standard way. Proposed tabletop experiments to test gravitationally induced entanglement between two mesoscopic masses each held in spatial superposition (proposals associated with Chiara Marletto and Vlatko Vedral, and independently Sougato Bose and collaborators) aim to probe exactly this: if gravity can mediate entanglement between two masses each in superposition, that is difficult to explain unless the gravitational field itself is carrying real, ontic quantum superposition — a striking, if indirect, argument bearing on the ontic status not just of matter's ψ but potentially of the gravitational field's quantum state too. These experiments are extraordinarily demanding, requiring superpositions of picogram-scale masses held coherent for seconds, and remain, honestly, some years to perhaps a couple of decades from feasibility — but they are being actively pursued in real laboratories, not merely proposed on paper.

What would it mean, ultimately, to call the question "settled"? A clean resolution would be an experiment ruling out every remaining psi-epistemic model without exception, including ones that abandon preparation independence — which looks close to impossible, since abandoning independence can always, in principle, be dressed up to mimic any observed statistics, the same stubborn flexibility that keeps superdeterminism alive against Bell's theorem. A messier but more realistic resolution is a slow, decades-long tightening of the noose: objective-collapse parameter space shrinking toward zero, epistemic-overlap bounds shrinking toward zero, gravitational entanglement experiments succeeding and favoring ontic gravitational superposition, until the epistemic alternatives that remain logically open become so contorted, fine-tuned, and explanatorily costly that the physics community treats them the way it now treats superdeterminism: not disproven, but abandoned by consensus as unworkable science.

It would be a mistake to think this is merely an argument about which philosophical adjective to attach to a piece of mathematics that works fine either way. The stakes are real, on at least three levels.

Scientifically, the answer bears directly on the measurement problem — arguably the single deepest unresolved issue in the foundations of physics — and on quantum cosmology's hardest question: what does it mean for the whole universe to have a wavefunction, if there is no external observer to update anyone's knowledge? An ontic reading demands we explain collapse, or eliminate it, as a real physical process; an epistemic reading demands we explain why nature's objective statistical regularities are so exquisitely law-like if ψ answers only to somebody's beliefs. Neither camp gets to declare victory and walk away from hard work.

Technologically, quantum information science has, in some ways, learned to work profitably in the space between the two positions — a point that touches the broader frontier of whether information itself should be regarded as physical. Quantum cryptography and quantum computing exploit precisely the features — no-cloning, monogamy of entanglement, the impossibility of extracting full classical information from a single copy of an unknown state — that PBR-style theorems trace back to the essentially ontic character of ψ. If ψ really does behave, for practical purposes, like a real physical resource rather than mere subjective opinion, that has already shaped how engineers build quantum key distribution networks and quantum random number generators; the metaphysics and the engineering are more entangled with each other than either camp always likes to admit.

Philosophically, and perhaps most personally, the question forces a decision about what kind of universe we live in at the most basic level available to us. If ψ is ontic, then reality genuinely contains entities — waves in absurdly high-dimensional configuration spaces, in most ontic pictures — utterly unlike anything in ordinary experience, and the classical world of definite tables and chairs is some kind of emergent, coarse-grained appearance riding atop a substrate almost unimaginably alien. If ψ is epistemic, then quantum mechanics is, at bottom, a theory of inference and information rather than a theory of stuff — closer in spirit to thermodynamics or Bayesian statistics than to Newtonian mechanics — and the deepest layer of physical reality, whatever it turns out to be, has not yet been written down by anyone. Both possibilities are, in their own way, unsettling; that is presumably why the argument has never gone quiet.

The dispute's earliest and most consequential skirmish was fought by two men who never fully reconciled. Niels Bohr, architect of the Copenhagen tradition, treated questions about ψ's "real" nature as ultimately meaningless outside the context of a specified experimental arrangement — famously resisting even the framing of ontic-versus-epistemic as a well-posed question at all. Einstein, unpersuaded, spent his final decades convinced the theory was an unfinished, statistical shadow of something more complete, a conviction that culminated in the 1935 EPR paper and in decades of subsequent correspondence with Born and Schrödinger, in which Einstein pressed exactly the "incomplete description" reading that later work would try to formalize.

physicist[Terry Rudolph|b. 1971|British-Australian theoretical physicist, co-author with Matthew Pusey and Jonathan Barrett of the 2012 no-go theorem now bearing their initials, which sharply restricts how epistemic any acceptable hidden-variable reading of ψ can be]

Christopher Fuchs and N. David Mermin have, since the early 2000s, pushed the most uncompromisingly epistemic reading in the form of QBism, insisting that quantum probabilities are personal Bayesian credences rather than objective features of the world — a position deliberately built to dissolve, rather than solve, the measurement problem. Rob Spekkens has done the opposite kind of work with equal rigor, constructing explicit toy models to map out exactly how much of quantum weirdness an honestly epistemic account can reproduce before it runs into trouble, turning what used to be hand-waving into precise, checkable mathematics on both sides of the debate.

:::history

• 1926 — Schrödinger's wave equation; Born's statistical interpretation of |ψ|²
• 1935 — Einstein–Podolsky–Rosen paper argues quantum mechanics is incomplete; Schrödinger's cat paper
• 1952 — Bohm revives and completes de Broglie's pilot-wave hidden-variable theory
• 1964 — Bell's theorem shows local hidden variables make testable, different predictions
• 2012 — Pusey–Barrett–Rudolph (PBR) theorem sharply constrains psi-epistemic models
• 2015 — Loophole-free Bell tests confirm quantum predictions, closing major local-realist escape routes

A century in, the argument has not gone stale so much as gotten sharper. What began as an almost aesthetic disagreement — does ψ feel more like a wave or more like a belief? — has been forged, by Bell, by PBR, by Spekkens, by the ongoing hunt for spontaneous collapse and gravitationally mediated entanglement, into a genuine research program with falsifiable edges and honest, unresolved gaps. That is real scientific progress, even without a final verdict. We have gone from "this is probably a question for philosophers" to "this is a question with theorems, loopholes in those theorems, and experiments that could, in specific and namable ways, narrow the loopholes further." Few interpretive questions in physics can claim that much movement.

What remains, stubbornly, is the residue the theorems cannot yet touch: preparation independence's uncertain status, the meaning of a single non-repeatable cosmic wavefunction, and the deep, still-unresolved question of whether "ontic" and "epistemic" are even the only two coherent options, or whether a genuinely third category — neither a plain fact nor a plain belief, something closer to a physical disposition or relation — is what the mathematics has been trying to tell us all along, in categories our ordinary metaphysics simply lacks the vocabulary to name.

• Does preparation independence hold as a fundamental principle of nature, or is it merely convenient and possibly false at some deep level?
• Can any experiment meaningfully probe ψ for a single, unrepeatable system — most urgently, the wavefunction of the universe itself?
• Will gravitationally mediated entanglement experiments, if they succeed, force an ontic reading not just of matter's ψ but of the gravitational field's quantum state?
• Is there a coherent third category — neither purely ontic nor purely epistemic — that better fits what the no-go theorems actually establish?
• Would a working theory of quantum gravity dissolve this question, transform it, or leave it exactly where it stands today?

Ask the tutor: If two physicists agree on every possible measurement outcome but disagree about whether ψ is real, what — if anything — could ever count as evidence that settles who is right?