What Is the Many-Worlds Interpretation of Quantum Mechanics?

In 2023, physicists reported a measurement of the electron's magnetic moment with a relative uncertainty of 0.13 parts per trillion, testing the Standard Model's prediction to about one part in a trillion. Quantum mechanics works. What it means is another matter. The theory says an electron can be in two states at once, yet every measurement finds only one. The Many-Worlds interpretation answers that puzzle by denying that anything ever disappears: every outcome happens, each in its own branch of reality.

Key takeaways

• The Many-Worlds interpretation (MWI) grew out of Hugh Everett III's 1957 "relative state" work and says the wave function never collapses.
• It uses only the Schrödinger equation. The collapse rule that textbooks add is dropped.
• Each measurement entangles the observer with the system, so every outcome survives in a branch where an observer sees it.
• Decoherence (the leaking of quantum information into the environment) is why the branches do not interfere in daily life.
• The hardest open problem is probability: why should outcomes carry weights that follow the Born rule, if all of them occur?
• No experiment can currently separate MWI from standard quantum mechanics, and surveys of physicists show no consensus.

What problem is Many-Worlds trying to solve?

Quantum mechanics describes a system with a wave function, a mathematical object that encodes every state the system might be found in and the amplitude for each. Between measurements, the wave function changes smoothly and deterministically according to the Schrödinger equation. A single electron's spin can sit in a superposition, a combination of "up" and "down" with specific amplitudes.

Then someone measures the spin. The apparatus shows "up" or it shows "down", never a blend. Textbook quantum mechanics handles this by adding a second rule: when a measurement happens, the wave function collapses to one definite outcome, with probabilities given by the squared amplitudes (the Born rule).

That second rule is where the trouble starts. The theory does not say what counts as a measurement. It does not say how big a device must be, whether a mouse can do it, or how long it takes. Physicists call this the measurement problem, and it has bothered the field for about a century.

Rule	What it says	Where it applies
Unitary evolution	The wave function changes smoothly by the Schrödinger equation	Always, between measurements
Collapse (projection)	The wave function jumps to one outcome, with Born-rule probabilities	"When a measurement occurs"

The collapse row is the vague one. The Copenhagen tradition associated with Niels Bohr and Werner Heisenberg handles it by treating the wave function as a tool for predicting what instruments record, and by placing a boundary between a quantum system and a classical measuring device. Critics ask where that boundary sits and why nature would respect it.

For a broader tour of the theory itself, Quantum Mechanics Explained covers the basics, and the Schrödinger's cat thought experiment shows the problem in its most famous form.

What did Everett actually propose in 1957?

Hugh Everett III was a graduate student at Princeton working under John Wheeler. According to the Stanford Encyclopedia of Philosophy, a long draft titled "Wave Mechanics Without Probability" went to Wheeler in January 1956. Wheeler pushed for revisions, and the much shorter version was accepted in March 1957 as a thesis titled "On the Foundations of Quantum Mechanics." A paper followed in Reviews of Modern Physics (volume 29, pages 454 to 462) under the title "'Relative State' Formulation of Quantum Mechanics." Everett always preferred the long version, and the encyclopedia notes that Bohr disapproved of his critical approach.

His proposal was radical because it was minimal. Keep the Schrödinger equation. Apply it to everything, including the measuring device and the physicist. Delete the collapse rule. Everett held that subsystems do not have independent states; in his words, "It is meaningless to ask the absolute state of a subsystem." An observer's experience is tied to the records that observer carries, and each branch of the global state contains an observer with a definite record.

Everett himself avoided talk of "worlds." The Stanford entry says Bryce DeWitt promoted the view under the name "many-worlds" in "Quantum Mechanics and Reality" in Physics Today (volume 23, 1970), which stressed physically splitting worlds. DeWitt presented it as the "EWG interpretation." Everett died in 1982.

How does a measurement turn into branches?

Imagine a spin measurement in steps:

1. An electron starts in a superposition of spin up and spin down.
2. A detector interacts with it. Under the Schrödinger equation, the detector does not choose. It becomes entangled with the electron: "up and detector reads up" plus "down and detector reads down."
3. You look at the detector. Now you are entangled too: "up, detector reads up, you see up" plus "down, detector reads down, you see down."
4. Each term in that sum contains a complete, self-consistent observer, who remembers a single outcome.

Nothing collapsed. The wave function just grew to include more of the world. A good analogy is a radio in a room where two stations broadcast at once. Both signals are physically present, but a receiver tuned to one station hears only that one. In MWI, you are a receiver inside one branch, and your memories contain only what happened in that branch.

Another way to see it is as a choose-your-own-adventure book read by a very strange reader. Every page that could follow your current page exists, and a version of the reader is on each of them. None of them sees the others, and each feels sure the page it landed on was the only one.

Does the universe really split every time anything happens?

Not quite, and this is the most common misreading. Branching is not triggered by decisions, consciousness, or human attention. It follows from interactions that spread a superposition into a larger system. A quantum event only produces branches that matter when it is amplified to a scale where its different outcomes become different macroscopic states.

The Stanford Encyclopedia of Philosophy is direct about the fuzziness: it states that "there is no precise definition of when the splitting occurs." That is not a fatal flaw. A comparable case is the idea of a wave on the sea. You can count waves, track them, and predict them, but no one can say exactly where one ends and the next begins. Branches work the same way: useful approximate patterns in the wave function, not sharp divisions written into the laws.

Why do you never see the other branches?

The answer is decoherence. When a quantum system interacts with a large, warm, messy environment such as air molecules, photons, and the rest of a detector, the information about its state leaks away into countless degrees of freedom. The different branches of the wave function stop being able to interfere with each other in any practical sense, and each one evolves as if the others were not there.

It happens fast. Max Tegmark's calculation of decoherence in the brain, published in Physical Review E in 2000, found timescales of roughly 10⁻¹³ to 10⁻²⁰ seconds, compared with neuron firing timescales of about 0.001 to 0.1 seconds. The point of that paper was a different debate, whether the brain is a quantum computer, but it illustrates how rapidly warm biological matter decoheres. A superposition of two brain states would not survive long enough to matter.

What is the preferred basis problem?

Mathematically, the same wave function can be split into branches in infinitely many ways, like cutting a pizza along different lines. The "preferred basis problem" asks why the branches we experience are the ones where the detector reads "up" or "down" rather than strange mixtures of both.

The Stanford Encyclopedia of Philosophy explains that the standard answer appeals to the locality of interactions: because the world's interactions are local, stable macroscopic states are localized ones, and decoherence picks them out. The entry says the problem "is no longer considered a serious objection," citing David Wallace's 2010 work, though it also notes that Meir Hemmo and Orly Shenker argued in 2022 that decoherence does not in fact solve it. So the question has a leading answer but is not entirely closed.

Why can't Many-Worlds just count the branches to get probabilities?

This is the deepest technical challenge, and it starts with an uncomfortable fact. In ordinary quantum mechanics, the Born rule is a postulate: the probability of an outcome is the squared amplitude. In MWI, all outcomes occur, so what does probability even mean?

Here is a worked example. Suppose an electron is prepared so that spin up has amplitude 0.8 and spin down has amplitude 0.6. Squaring them gives 0.64 and 0.36, which sum to 1. Laboratory results confirm this: over many runs, spin up shows up in about 64 percent of trials.

A naive MWI count would say the measurement creates two branches, so each outcome should be 50 percent likely. That conflicts with experiment. The Stanford Encyclopedia of Philosophy records that deriving probability from the number of worlds "leads to the wrong predictions." So MWI needs another account of what the weights mean.

What are the main proposed answers?

Approach	Core idea	Main objections
Decision theory (Deutsch 1999, Wallace)	A rational agent who accepts MWI should bet according to the squared amplitudes	Critics argue the assumptions are circular or not compelling (Barnum et al. 2000, Baker 2007, Albert 2010, Price 2010)
Envariance (Zurek 2005)	A symmetry of entangled states, in which a change on one part is undone by a matching change on the other, fixes the probabilities	See the Stanford entry for the critiques
Measure of existence (Vaidman)	Each world has a "measure of existence" equal to its squared amplitude, instead of being counted	Albert objects that the self-locating probability arises "too late"
Self-locating uncertainty (Sebens and Carroll)	Before you look, you are uncertain which branch you are in	Kent (2015) and McQueen and Vaidman (2019) argue it begins with a question that is meaningless before branching

David Deutsch's 1999 paper, "Quantum Theory of Probability and Decisions," argues that the probability axiom is unnecessary, because all practical consequences follow from the non-probabilistic axioms of quantum theory plus the non-probabilistic part of classical decision theory. David Wallace later axiomatized the approach around three assumptions: symmetry of unitary quantum mechanics, consistency of preferences over time, and indifference to fine-grained branching.

The Stanford entry describes the incoherence problem, raised by Wallace in 2003, as "arguably the most serious difficulty." Defenders call the decision-theoretic proofs a genuine derivation. Skeptics say they smuggle in the very idea of probability they claim to explain. Neither side has won.

What are the other objections?

Ontological extravagance. A branching multiverse seems like a heavy price. The Stanford entry says critics call the proliferation of unobservable worlds an extreme violation of Ockham's razor. Defenders reply that MWI removes the collapse postulate, which they regard as the most problematic law in the theory. The worlds are not added by hand; they are what the equations already contain. Hemmo and Shenker agree MWI is economical but argue this is irrelevant because the postulates alone cannot explain experience.

The wrong kind of object. Tim Maudlin argues the wave function lives in a high-dimensional configuration space (3N dimensions for N particles), not in ordinary spacetime, which would make it the wrong type of object to describe the world we see. The Stanford entry credits this argument to Maudlin (2010).

Testability. No experiment currently favors collapse over MWI. The Stanford entry notes that some collapse models predict effects such as energy non-conservation, and some have been ruled out by experiments such as those of Vinante and colleagues in 2020. A decisive test would need interference involving a macroscopic body, beyond current technology. That is why many call MWI an interpretation rather than a separate theory. If collapse models were confirmed, however, MWI would be false, so the question is not entirely empty.

What happens to Schrödinger's cat in Many-Worlds?

Erwin Schrödinger's cat was meant as a reductio: if a quantum event can put a flask of poison in superposition, then a cat is both alive and dead, which seemed absurd. Collapse theories answer by saying that something, whether the detector, the cat, or the observer, triggers a jump to one outcome. They disagree about which.

Many-Worlds gives the plainest reply. The radioactive atom, the detector, the poison, the cat, and the person who opens the box all become entangled. The wave function contains a branch with a living cat and a relieved observer, and a branch with a dead cat and a sad one. Neither the cat nor the observer experiences a mixture, because within each branch the records are definite. Decoherence makes the two branches practically independent almost at once, long before anyone lifts the lid.

This is also a reminder of what the thought experiment was for. Schrödinger meant to expose a gap in the theory, not to endorse a half-dead animal. Many-Worlds closes the gap by accepting the strange description rather than adding a rule that removes it.

A short timeline of the idea

Date	Event
January 1956	Everett gives Wheeler the long draft, "Wave Mechanics Without Probability"
March 1957	Shorter thesis, "On the Foundations of Quantum Mechanics," is accepted at Princeton
1957	"'Relative State' Formulation of Quantum Mechanics" appears in Reviews of Modern Physics 29:454-462
1970	DeWitt's Physics Today article presents a splitting-worlds version
1999	Deutsch proposes a decision-theoretic account of probability
2000	Tegmark publishes decoherence timescales for brain processes in Physical Review E
2020	Bong and colleagues publish a Wigner's-friend no-go theorem in Nature Physics
December 9, 2024	Google announces the 105-qubit Willow chip, with a remark about parallel universes
2025	Nature survey finds 15 percent of respondents favor Many Worlds, per The Quantum Insider's summary

The sequence runs from a graduate thesis to decoherence calculations, quantum-computing announcements, and large surveys. The core questions remain unsettled.

How does Many-Worlds compare with the other interpretations?

Several interpretations of quantum mechanics are in active use. Most reproduce the same predictions for experiments we can run today. They disagree about what is real.

Interpretation	What happens in a measurement	Status of the wave function	Differs experimentally?
Copenhagen	Collapse upon interaction with a classical device	Tool for predicting results (in many versions)	No
Many-Worlds	No collapse; all outcomes persist in branches	Fully real	No
de Broglie-Bohm (pilot wave)	Particles have definite positions, guided by the wave	Real, guiding field	No, by design
Objective collapse (GRW, Penrose)	Spontaneous collapse added to the dynamics	Real	Yes, in principle
QBism and epistemic views	Updating of an agent's beliefs	Encodes information or belief	No
Relational quantum mechanics	States are relative to systems	Relative	No

The pilot-wave approach is deterministic but requires nonlocal influences, which is compatible with the results of Bell tests. John Bell, who proved the theorem named for him, spent much of his career arguing that the measurement problem was real; his life and work are covered in John Bell: the physicist who proved Einstein wrong about quantum reality, and the experiments that tested his theorem are described in the Aspect Bell test.

What do physicists actually think?

Opinion polls are not evidence about nature, but they tell us how the field regards the ideas. Two stand out.

In a 2011 poll of 33 participants at a conference on the foundations of quantum mechanics in Traunkirchen, Austria, reported by Maximilian Schlosshauer, Johannes Kofler, and Anton Zeilinger in 2013, the question "What is your favorite interpretation of quantum mechanics?" got these answers: Copenhagen 42 percent, information-based 24 percent, Everett (many worlds and/or many minds) 18 percent, objective collapse 9 percent. Respondents could pick more than one, so the figures add to more than 100 percent. The sample was tiny and self-selected.

A far larger survey came in 2025, the centennial year of the theory's modern formulation. According to The Quantum Insider's summary of a Nature survey, more than 1,100 researchers responded out of more than 15,000 contacted. Copenhagen led at 36 percent, Many Worlds got 15 percent, Bohmian mechanics 7 percent, and epistemic views about 17 percent. Only 24 percent said they thought their favorite interpretation was correct. Others called it useful or the least problematic option. Of the respondents, 86 percent said interpretations are a valuable pursuit, and 58 percent said empirical tests could help settle the debate. These figures come from secondary coverage; the Nature article itself was not accessible for this review.

Two things follow. Many-Worlds is a minority view even among specialists, but it is a substantial one. And almost no one is certain of their own answer.

Does quantum computing show that parallel universes are real?

David Deutsch, a pioneer of quantum computing and a leading MWI advocate, has long argued that quantum computers draw on many branches at once. In a December 9, 2024 post announcing the 105-qubit Willow chip, Google's Hartmut Neven wrote that its result "lends credence to the notion that quantum computation occurs in many parallel universes," and credited the idea to Deutsch. The chip finished a random circuit sampling task in under five minutes that the post said would take one of the fastest supercomputers about 10 septillion (10²⁵) years.

That sentence is a statement of preference, not a proof. The interpretations that keep the standard equations predict the same outcome for that computation, because they agree on the mathematics. A quantum computer's speedup is evidence that quantum mechanics is hard to simulate classically. It does not tell you whether the underlying story involves branches, hidden variables, or something else. For how such machines work in practice, see how quantum computers actually work.

Can experiments test what counts as an observer?

A recent line of work probes the same territory from another side. In the "Wigner's friend" thought experiment, an observer (the friend) measures a quantum system inside a sealed lab, while a second observer outside treats the whole lab as one quantum system. In 2020, Kok-Wei Bong and colleagues reported in Nature Physics a strong no-go theorem and a proof-of-principle test. They argue that if quantum evolution can be controlled at the scale of an observer, one of three assumptions must fail: no superdeterminism, locality, or the absoluteness of observed events.

The experiment used a photon's path as a stand-in for the friend, so it does not show how a real human observer would behave. Still, Everett's relative-state idea is one way to give up that third assumption, since he held that an observer's records are relative to a branch. These experiments do not select MWI. They map which assumptions any interpretation has to give up. Another experiment that probes what measurement does to interference is covered in the quantum eraser experiment.

How is this different from the cosmological multiverse?

The word "multiverse" covers several unrelated ideas. Cosmological multiverses involve other regions of space, possibly with different physical constants. Many-Worlds branches are not located anywhere in space. They are components of one wave function, and they coexist in the same spacetime, the way two radio signals coexist in the same room.

Some physicists have tried to connect the concepts. Most keep them separate, because the evidence and the reasoning are different. For the cosmological version, see multiverse theory explained, and for a hypothetical look at what parallel universes would imply, what if parallel universes are real.

What do people commonly get wrong about Many-Worlds?

"The universe splits every time you make a choice." Branching follows physical interactions that amplify quantum differences, not human choices. Making a decision does not carry special weight.

"Anything imaginable happens." Only outcomes with non-zero amplitude in the wave function appear, and different branches have different weights. A branch in which you win the lottery by a quantum coin flip exists only if the physics gives the amplitude.

"You can visit or signal other branches." Decoherence makes them effectively isolated. There is no mechanism for travel, and no observation inside one branch can reveal the others directly. Because branches are not places in space, the concept is unrelated to speed limits, a topic explored in what moves faster than light.

"It is pure science fiction." It is a serious interpretation, but it is also an interpretation, not a separate theory. It adds no new equations. What it adds is a claim about what the existing equations describe.

"Consciousness is special in MWI." Not in the formalism. Any physical interaction that entangles systems can create branches. Some early discussions treated consciousness as special, but contemporary versions such as Wallace's do not require it.

What does Many-Worlds mean for spacetime and gravity?

Many-Worlds treats the wave function as the fundamental object, which raises a question about what space and time are. As noted above, the wave function lives in a configuration space with 3N dimensions for N particles, so the familiar three-dimensional world has to emerge from it. That makes MWI relevant to attempts to unify quantum theory and gravity. See what quantum gravity is and what spacetime is for the surrounding questions. At present, no experiment links the interpretation question to a specific gravitational prediction.

FAQ

Q: Are the parallel universes in MWI the same as the multiverse in cosmology? A: No. The cosmological multiverse refers to regions of space beyond our observable universe. The quantum branches in MWI are components of a single wave function that do not exist as separate places in space. Some physicists try to unify the ideas, but most treat them as separate questions.

Q: Could you ever visit a parallel MWI universe? A: No. Decoherence makes branches effectively isolated, and there is no physical pathway between them. No observation made inside one branch can reveal the others directly.

Q: Did Einstein support the Many-Worlds interpretation? A: No. Einstein died in 1955, before Everett's 1957 thesis. He was uncomfortable with the randomness of Copenhagen quantum mechanics, but whether he would have preferred Many-Worlds is unknown.

Q: Is consciousness special in Many-Worlds? A: No, not in the mathematical formalism. Any interaction that entangles a system with its environment produces branching, with or without a conscious observer.

Q: Why do I only experience one outcome if all outcomes happen? A: Because you are also in superposition. After a measurement, there are versions of you in each branch, and each remembers only its own outcome. Decoherence prevents them from interfering or communicating.

Q: Which interpretation is correct? A: Nobody knows. Surveys show no consensus: in the 2025 Nature survey as reported by The Quantum Insider, Copenhagen led at 36 percent and Many Worlds drew 15 percent, and only 24 percent of respondents thought their favored view was correct. Decisive tests would need new experiments, such as macroscopic interference that rules collapse models in or out.

The takeaway

The Many-Worlds interpretation is a bet on taking the equations at face value. If the Schrödinger equation holds for electrons, detectors, and people alike, then superpositions spread outward and never vanish, and what we call collapse is how one branch looks from the inside. The price is an enormous branching structure, plus a hard unsolved question about why branch weights should guide our expectations.

That price is why the debate lasts. Many-Worlds is not proven, and it is not refuted. It may be the plainest reading of the theory, or it may be a mathematical shadow of something simpler. Until an experiment separates it from its rivals, the honest position is that it is a serious candidate for what quantum mechanics means.

Sources

• Stanford Encyclopedia of Philosophy, Many-Worlds Interpretation of Quantum Mechanics and Everett's Relative-State Formulation of Quantum Mechanics
• Fan, Myers, Sukra, and Gabrielse, Measurement of the Electron Magnetic Moment, Physical Review Letters 130, 071801 (2023)
• Deutsch, Quantum Theory of Probability and Decisions (1999)
• Tegmark, The importance of quantum decoherence in brain processes, Physical Review E 61, 4194 (2000)
• Schlosshauer, Kofler, and Zeilinger, A Snapshot of Foundational Attitudes Toward Quantum Mechanics (2013)
• The Quantum Insider, A Century Into Quantum Mechanics, Physicists Still Can't Agree What It Means, Nature Survey Shows (2025)
• Google, Meet Willow, our state-of-the-art quantum chip (December 9, 2024)
• Bong et al., A strong no-go theorem on the Wigner's friend paradox, Nature Physics (2020)

Related Reading

• Quantum Entanglement: What It Is and What It Isn't
• The Double-Slit Experiment: How Light Proved the Universe Is Weird
• Erwin Schrödinger: Wave Mechanics, the Quantum Cat, and a Complicated Life
• The Bohr-Einstein Debates: The Thought Experiments That Shaped Quantum Physics
• Quantum Mechanics Foundations Explained: Wave, Uncertainty, and Entanglement