Interpretations, Decoherence & Schrödinger's Cat

Late in 1935, months after Einstein, Podolsky and Rosen accused quantum mechanics of being incomplete, Erwin Schrödinger published an essay, 'The Present Situation in Quantum Mechanics'. It put a name to entanglement and, in one deadpan paragraph, introduced physics' most famous animal. A cat is shut in a steel chamber with what Schrödinger called an infernal device: a Geiger counter holding so little radioactive material that within an hour perhaps one atom decays, or, with equal probability, none does. A decay trips the counter, which releases a hammer that shatters a flask of hydrocyanic acid, killing the cat.

Now simplify the sample to a single atom with a one-hour half-life, and apply the Schrödinger equation to everything in the chamber. The equation is linear: if two states are possible evolutions, so is any superposition of them. After an hour the atom is in an equal superposition of decayed and undecayed, and because counter, hammer, flask and cat form a causal chain, linearity carries that superposition all the way up. The joint state becomes |undecayed⟩|alive⟩ + |decayed⟩|dead⟩, divided by √2: an entangled state like those of Chapter 20. The wavefunction, Schrödinger wrote, would contain 'the living and dead cat (pardon the expression) mixed or smeared out in equal parts'.

Schrödinger was not announcing that cats can be half dead. He was building a reductio ad absurdum, an argument that derives a ridiculous conclusion to discredit a premise: here, that the wavefunction is a literal, complete portrait of reality. A blurred photograph of a sharp object, he observed, is not the same as a crisp photograph of a fog bank, and for the cat the wavefunction resembles the blurred photograph. The modern question is sharper: what physical story turns that entangled state into the single, unsmudged cat we always find?

The cat wraps a precise difficulty called the measurement problem. Textbook quantum mechanics, as Chapter 15 showed, runs on two rules. Rule one: an isolated system evolves smoothly and deterministically by the Schrödinger equation. Rule two: on measurement the state collapses at random to one outcome, with probability equal to the squared magnitude of that outcome's amplitude (the Born rule). But a measuring device is made of atoms too. If rule one governs it, counter and cat end in a superposition and no jump ever happens; if rule two takes over, we need a physical definition of 'measurement' that says when, and the textbooks never supply one.

In 1995 the philosopher Tim Maudlin sharpened this into a trilemma of three claims that cannot all be true. First, the wavefunction is a complete description of a physical system. Second, it always evolves by the linear Schrödinger equation. Third, every measurement has a single, definite outcome. Accept any two and the cat argument forces you to reject the third. An interpretation of quantum mechanics, a proposed account of what the mathematics says about the world, can be classified by the claim it gives up, or by how it reinterprets the wavefunction. That is our map.

The oldest answer belongs to the Copenhagen family, the overlapping views of Niels Bohr, Werner Heisenberg and their circle from the late 1920s. It was never one doctrine, and the name itself apparently first appears in a 1955 essay by Heisenberg. Bohr insisted that quantum phenomena can be described only through whole experimental arrangements, themselves described in classical language: a counter clicks or it does not. His central idea, complementarity, first presented at Como in 1927, adds that mutually exclusive arrangements reveal mutually exclusive aspects, such as sharp position and sharp momentum, both needed for a full account. For Bohr the wavefunction was a symbolic tool for predicting outcomes, not a picture of a thing, and he never spoke of a collapse of the wave packet.

The textbook version adds a pragmatic rule: divide the world into a quantum system and a classically described apparatus, separated by a movable boundary called the Heisenberg cut. John von Neumann formalized this in 1932 with his projection postulate, which on measurement replaces the state by the state matching the result. Putting the cut at the counter, the cat or the lab notebook gives identical predictions. Read in Bohr's spirit, the cat poses no riddle: the counter's irreversible avalanche of ions is already a measurement, so the cat is alive or dead long before anyone lifts the lid.

The price is vagueness. Copenhagen never says physically what makes something classical, or exactly when rule two takes over. John Bell's 1990 essay 'Against measurement' protested that a fundamental theory should not lean on a word it cannot define. Defenders reply that physics concerns what can be said about experiments, and in daily practice the attitude David Mermin summed up as 'shut up and calculate' serves most physicists well.

In 1957 Hugh Everett, a doctoral student of John Wheeler at Princeton, proposed the boldest repair: delete rule two. In his 'relative state' formulation the wavefunction of the whole universe always obeys the Schrödinger equation, and nothing ever collapses. A measurement is just an entangling interaction. After an hour the box holds one branch in which the atom is intact and the cat alive, and another in which the atom has decayed and the cat is dead, each with weight one half. Bryce DeWitt popularized it in 1970 as 'many worlds'.

Each branch contains a version of the observer who sees one definite cat: every outcome happens, though each copy of you sees just one. The branches are not universes stamped out by a cosmic photocopier but components of one vast wavefunction that have stopped interfering, through the process of Part 7. Supporters prize the economy: no extra particles, no extra dynamics, no special role for observers and, they argue, no action at a distance, since Bell's theorem assumes single outcomes.

The costs are an extravagant ontology and a puzzle about probability: if every outcome occurs, what does it mean that one has probability 0.9? Everettians identify probability with each branch's squared amplitude, or weight, and David Deutsch in 1999 and David Wallace later argued that a rational agent in a branching world should bet exactly by the Born rule. Whether those arguments succeed is still fiercely contested.

The second route keeps single outcomes and the Schrödinger equation by denying that the wavefunction is the whole story. In pilot-wave theory, sketched by Louis de Broglie at the 1927 Solvay Conference and rediscovered and completed by David Bohm in 1952, every particle has a definite position at every moment. The wavefunction still obeys Schrödinger's equation, but its job is to steer: a guiding equation sets each particle's velocity from the gradient of the wavefunction's phase. The theory is completely deterministic.

Randomness comes only from ignorance of starting positions. If positions in an ensemble begin distributed as |ψ|², a condition called quantum equilibrium, they stay so distributed, and every Born-rule prediction follows. In the double slit each electron passes through one slit while its guiding wave passes through both and sculpts the fringes. In the box the cat's atoms always have positions, so the cat is simply alive or dead; the unoccupied part of the wave becomes an 'empty wave' that, after decoherence, never again affects the particles.

The theory is explicitly nonlocal. For several particles the wavefunction lives in configuration space, the space of all the particles' positions taken together, so when particles are entangled one particle's velocity depends instantly on where the others are, however distant. That is exactly the price Bell's theorem demands of hidden variables, and Bell openly admired Bohm's theory. Quantum equilibrium ensures that the nonlocality can never carry a signal, but it sits awkwardly with relativity: simple relativistic versions need a hidden preferred notion of simultaneity.

The third route changes the Schrödinger equation itself. In 1986 GianCarlo Ghirardi, Alberto Rimini and Tullio Weber proposed that every particle occasionally suffers a spontaneous, random 'hit' that localizes it to about 10⁻⁷ metres. For one nucleon, a proton or neutron, the rate is about 10⁻¹⁶ per second, a hit every 300 million years or so, so atoms behave just as standard theory predicts. But in a macroscopic superposition the particles are entangled, so a hit on any one displaced particle collapses the whole state, and with some 10²³ nucleons the hits arrive every 10⁻⁷ seconds. The cat is collapsed long before it could be smeared.

Lajos Diósi in 1987 and, independently, Roger Penrose from 1989 onward argued that gravity does the collapsing: a superposition of two different arrangements of mass should decay in a time of roughly ħ divided by the gravitational self-energy of their difference. Collapse models are rival theories, not interpretations: they modify the dynamics, and their random hits jiggle particles, so charged ones should glow faintly. In 2020 Sandro Donadi, Angelo Bassi and colleagues looked for that glow with a germanium detector under the Gran Sasso massif in Italy and ruled out the parameter-free Diósi-Penrose model. Interference of ever heavier particles, most recently sodium clusters above 170,000 atomic mass units (Vienna, January 2026), squeezes the models from another side; yet as of the Stanford Encyclopedia's 2025 review, no experiment had reached the original GRW values.

Two further families reinterpret the wavefunction instead. QBism, pronounced like 'cubism' and grown from work begun in 2002 by Carlton Caves, Christopher Fuchs and Rüdiger Schack, treats a quantum state as an agent's personal degrees of belief about their own future experiences; collapse is that agent updating on new experience, so nothing physical happens to the cat when you open the box. Carlo Rovelli's relational quantum mechanics (1996) holds that properties exist only relative to other systems: relative to the cat the outcome is definite, while relative to a physicist outside the sealed box, cat and atom are rightly described as entangled. Critics ask what, then, the world is like in itself.

Every interpretation must explain one stubborn fact: we never see interference between macroscopically different states. The explanation, called decoherence, was founded by H. Dieter Zeh in 1970 and developed by Wojciech Zurek in the early 1980s and by Erich Joos and Zeh in 1985. No real object is isolated: air molecules strike it, photons scatter from it, and it radiates heat. Each encounter leaves the environment slightly different depending on the object's state, so the environment becomes entangled with the object and carries a record of it, spread across countless particles.

That record kills interference, because two alternatives can interfere fully only if they leave everything else in the same final state. Physicists describe the object alone by its reduced density matrix, whose diagonal entries are the probabilities of each alternative and whose off-diagonal entries, called coherences, carry the ability to interfere. Entanglement multiplies each coherence by the overlap γ between the environment states recording the two alternatives. If each passing particle carries a little information, its γ is slightly below one, and N independent particles multiply the coherence by γ raised to the power N, which plunges exponentially.

A dust grain ten micrometres across in still air is struck by about 10¹⁸ gas molecules every second, each recording its position, so a superposition of two locations decoheres in about 10⁻¹⁸ seconds. Zurek estimated that for a one-gram object at room temperature, superposed over one centimetre, decoherence beats ordinary frictional relaxation by a factor of order 10⁴⁰. A warm, breathing cat has no chance of showing interference between alive and dead.

Experiments have watched it happen. In 1996 Serge Haroche's group in Paris put a few-photon microwave field into a superposition of two phases inside a superconducting cavity, a miniature Schrödinger cat, and saw its coherence fade faster the more different the components were; Haroche shared the 2012 Nobel Prize in Physics. In 2003 Klaus Hornberger, Markus Arndt, Anton Zeilinger and colleagues in Vienna sent C₇₀ molecules through an interferometer while admitting traces of gas: the fringe visibility fell exponentially with pressure, because a single collision erased a molecule's interference.

Decoherence is mainstream physics, not an interpretation. Yet the total state of cat, atom and environment is still a superposition evolving by the Schrödinger equation; nothing collapsed. The cat's reduced density matrix merely looks like a coin toss, alive or dead with probability one half each, once we ignore the environment. Philosophers call this an improper mixture: it comes from setting aside part of an entangled whole, not from ignorance of an outcome that has already happened. Decoherence explains why the two cats never interfere. It does not say which cat you see, or why you see only one.

Each interpretation uses decoherence differently. For many-worlds it defines the branches. For pilot-wave theory it stops empty waves from ever interfering again with occupied ones. For Copenhagen it explains why the cut can sit anywhere above the microscopic scale without changing a prediction. For collapse models it is the noise that must be suppressed before a genuine collapse could be detected.

For every experiment yet performed, all of these interpretations give identical predictions, because they share the same mathematics, and collapse models deviate by amounts too small to have been seen. Which story is true is an open question. When Nature surveyed physicists in 2025, the centenary of quantum mechanics, its headline said they disagree wildly about what the theory says about reality. I'm mad, not wrong: the experiments are settled; the story is not.