Concept

Many-Worlds Interpretation

conceptphysicsquantum-mechanicsphilosophy-of-science
Teach me this →

Many-Worlds Interpretation

The claim, due to Hugh Everett III (1957), that quantum mechanics needs no collapse postulate — the Schrödinger equation applies everywhere and at all times, and what we call a measurement outcome is simply the observer becoming entangled with the system, producing a superposition in which every outcome exists in a separate branch. Each branch is a ‘world’: fully real, non-interacting with the others, and containing observers who each see a definite result. The interpretation is also called the Everett interpretation, or Everettian quantum mechanics.

Sean Carroll is its most prominent current advocate, arguing that many-worlds is not a baroque addition to quantum mechanics but its minimal reading — what you get when you take the equations seriously and refuse to add extra postulates to suppress their implications.

The wiki’s second Everettian voice, Simon Saunders, shares the parsimony case but reads the formalism from a decoherent-histories starting point, derives the Born rule by a different route, and adds an argument about locality — set out under A second Everettian below.

The structure of the argument

Quantum mechanics begins with the wave function (or, more precisely, a vector in Hilbert space) and the Schrödinger equation governing how it evolves. Both are uncontroversial: every interpretation of quantum mechanics starts here. The disagreement concerns what to do when the equations predict that a measurement leaves the wave function in a superposition — for example, an electron in a box that is simultaneously in box A and box B, combined with an observer who has simultaneously recorded ‘box A’ and ‘box B’.

Copenhagen-family interpretations add a further postulate: at the moment of measurement, the wave function collapses to one branch. The probability of each outcome is given by the Born rule (the squared amplitude of each branch). This resolves the apparent contradiction with experience — observers always see definite outcomes — but at a cost: the collapse rule is undefined. What constitutes a measurement? What makes an observer? No precise answer has been given.

Everett’s move: decline to add the collapse postulate. Accept that the superposition is real, that both branches exist, and that the observer in each branch sees a definite result. ‘You are not the superposition of both outcomes — you are one or the other. And there is another observer who shares your past who is in the other branch.’ The wave function of the universe branches; ordinary experience is recovered by locating yourself correctly within it.

The appeal: parsimony

The case for many-worlds is not that it is comfortable but that it is simple. Counting the axiomatic commitments required:

  • Copenhagen: wave function + Schrödinger equation + collapse postulate + Born rule for collapse probabilities + a definition of measurement (which Copenhagen has never supplied).
  • Everett: wave function + Schrödinger equation. Full stop.

Fewer axioms is a recognised scientific virtue. The ‘many worlds’ are not additional ontology smuggled in — they are a consequence of the single wave function that every other interpretation also requires. As Carroll puts it, the extra worlds cost nothing: they are already implied by the equations everyone accepts.

The open questions

Many-worlds is parsimonious in its postulates but demands philosophical work to connect the formalism to experience. Carroll identifies two central open problems:

The Born rule problem. In Everett, all outcomes occur, so asking ‘how often does outcome X happen?’ is ill-formed. The challenge is to derive the Born rule — the squared-amplitude probability weights — from the structure of the theory alone, without simply postulating it. Carroll’s favoured approach reframes the question as self-locating probability: given that you are about to branch into multiple copies, one in each outcome, how should a rational agent apportion credence across them? He argues that branch-counting alone gives the wrong answer, and that a decision-theoretic argument (rational agents who care about all future selves should weight branches by squared amplitude) recovers the Born rule. This is contested.

The problem of structure. Classical space, separable particles, and local interactions are nowhere in the bare Hilbert-space formalism. Showing that they emerge — that the unique correct way to carve Hilbert space into subsystems recovers the space we observe — is an active research programme. Carroll and collaborators argue that most Hamiltonians admit no local subdivision at all; the cases where locality holds are special and rare, suggesting that the locality of our universe is emergent and possibly only approximate.

A second Everettian: Saunders on the Born rule and locality

Simon Saunders, an Oxford philosopher of physics, defends Everett from a decoherent-histories starting point and diverges from the account above on two substantive points.

The first concerns the probability argument. The decision-theoretic derivation of the Born rule described above originates with the Oxford school — David Deutsch’s argument, developed and defended through the 2000s by David Wallace and Saunders — rather than with its later advocates. Saunders’ own current position has moved past decision theory. He derives the Born rule instead from a single physical postulate that extends Bell-style locality to probability itself: a physically allowed action on a system Y cannot change the probability of a disjoint system X. Under unitary quantum mechanics, where every physical change is unitary, this forces equal-amplitude branches to carry equal probability — a strict derivation rather than a postulate. But the method, which he calls finite frequentism, yields probabilities as intervals rather than sharp numbers: an irreducible grey zone remains, and the most extreme low-amplitude, Born-rule-violating branches (he compares them to Boltzmann-brain scenarios) can never be assigned a definite probability by any finite analysis — ‘you will not even see the probability’ for them.

The second concerns locality. Saunders argues — the boldest claim of Simon Saunders on the Nature of Time, Many-Worlds, and Bell's Inequalities — that violations of Bell’s inequalities are evidence for many-worlds rather than against locality. Two correlated observers, each strictly confined to their own branch, are so tightly correlated that they can be treated as a single ‘extended observer’ inhabiting a common branch; Saunders suggests this branching-correlation structure is what underlies the mere appearance of Bell non-locality in entangled-particle experiments. He states the thesis but, running short of time in the conversation, defers its full derivation — so within the wiki it stands as an open argument, not a completed one.

Saunders is also wary of Everettian pictures that bypass decoherence theory, singling out the reading on which worlds correspond one-to-one with the discrete eigenspaces of the energy operator as ‘very different from anything that I have gathered from Everett’ (he does not name whose view he means).

Where mainstream views differ

Most physicists, when surveyed, align with Copenhagen-family views (roughly one-third) or remain agnostic. The main objections to Everett:

  • The Born rule is not derived, it is assumed. Decision-theoretic arguments presuppose that one should care about all branches, which itself imports a non-trivial assumption about personal identity across branching.
  • The preferred basis problem. The Schrödinger equation alone does not specify which observable is ‘classical’ — it does not explain why branches decohere along position rather than momentum or some other basis. Decoherence theory (Zurek, Joos) partially addresses this: interaction with the environment picks out a preferred basis. But the full derivation from first principles is not complete.
  • Metaphysical extravagance. Even if many-worlds requires fewer axioms, the ontological commitment to infinitely many unobservable branches strikes many physicists as too high a price.

Carroll’s response: open questions are opportunities, not refutations. If no acceptable derivation of the Born rule or emergence of space is found, that would be a reason to abandon Everett. He does not think we are there.

In the wiki