Bell's theorem and quantum reality

The interpretations zoo

Bell tests reveal a hard fact: when detector choices are independent, no complete explanation based only on each nearby setting and the shared past can reproduce nature's correlations. Quantum mechanics gets them right. The mystery is what kind of world does that.

Run the Bell experiment See the exit doors
Locality means that, after accounting for the shared past, a distant choice or result should not be needed to explain what happens here. Bell tests defeat that kind of local explanation, but they still do not let anyone send a faster-than-light message.
Distant detector ledger individual runs look random
85.4% quantum win rate
across four setting pairs
chanceuniform-setting local ceiling: 75%perfect
no usable message crosses the gap the pattern appears only after comparison
01 / THE MYSTERY

Local common causes are not enough

A Bell-local theory says each detector answers using only its nearby setting and information carried from the shared past.

Matching instruction cards can fake some agreement. The shock is that, across four possible pairs of settings, nature wins a correlation game no such local strategy can win often enough.

01

Prepare a pair

A source emits two entangled photons toward laboratories so far apart that each measurement finishes before light could cross between them.

02

Choose settings late

Alice and Bob uniformly and independently choose one of two detector orientations only after the pair is already in flight.

03

Record one bit

Each side gets an unpredictable 0 or 1. Looking at one laboratory alone reveals no pattern and carries no remote message.

04

Compare later

Only after ordinary, slower-than-light communication brings the ledgers together does the forbidden aggregate pattern appear.

Bell's correlation game

Live simulation with uniformly random, independent settings

Pairs leave a central source. Alice and Bob choose one of two detector angles, receive random binary outcomes, and add the run to the aggregate statistics. The animation never shows a signal traveling between the detectors.

RunSettingsOutcomesRule
Local theory's ceiling With four setting pairs sampled uniformly, at most three parity rules can be pre-satisfied.
S ≤ 2   vs   S = 2√2
Quantum prediction Experiments approach 2.828 while preserving no-signaling.
02 / THE EXIT DOORS

What has to give?

Bell rules out one comfortable combination, not every possible worldview.

You cannot keep independently chosen settings, a complete account of the shared past, and purely local responses while reproducing the observed correlations. Every interpretation changes at least one part of that picture.

Bell did not prove that "nothing is real." Modern Bell arguments do not need a separate assumption called realism. The sharper question is whether a complete account of the shared past can make both detectors respond locally.

Exit 01Bell core

Must causes stay local?

Formal term: Bell local causality. Once the shared past is fully described, Alice's result should need only Alice's setting, and Bob's result should need only Bob's setting.

The remote knob: Bob's setting should not alter Alice's underlying probabilities. This is parameter independence. The remote result: learning Bob's outcome should add nothing once the shared past and both settings are known. This is outcome independence.

Show the mathematical versionP(A,B|a,b,λ) = P(A|a,λ) P(B|b,λ)
Exit 02Bell core

Were the detector choices truly independent?

Formal term: measurement independence. The hidden state of the pair should not already be correlated with which settings Alice and Bob later choose.

Show the mathematical versionρ(λ|a,b) = ρ(λ)
Exit 03Interpretive extension

Is there one result for everyone?

Formal idea: a single global outcome. Some interpretations expect every run to produce one observer-independent result pair. Modern Bell arguments do not assume this separately, but interpretations disagree sharply about it.

Show the compact versionone run → one absolute fact pair
Exit 04Causal extension

Can the future help shape the past?

Formal idea: a forward-only causal arrow. Ordinary explanations let settings influence later events, not variables in their own past. Time-symmetric theories relax that rule.

Show the compact versionpast → settings → outcomes
Exit 05Dynamics

Does the wave equation ever break?

Formal term: universal unitary evolution. Many-Worlds says the Schrödinger equation always applies. Objective-collapse theories add stochastic localization dynamics: rare discrete hits in GRW, continuous noise in CSL.

Show the mathematical questioniℏ d|ψ⟩/dt = H|ψ⟩   always?
Exit 06Meaning

Is the wavefunction a thing in the world?

Formal idea: an observer-independent quantum state. Some views treat the wavefunction as physical. Others treat it as relational information or an agent's probability assignment.

Show the compact questionstate of the world   or   state for a system?

The last four are not all premises in the same Bell derivation. They are broader commitments that determine how an interpretation reads Bell's result. Several interpretations use more than one exit.

03 / THE ZOO

Nine serious ways to pay the bill

Most animals reinterpret the same Bell data; objective collapse also changes the dynamics.

Familiar names come first; formal names and qualifications stay close by. Each card explains what the view changes, what it says exists, and whether it predicts new experimental results.

9 animals
Copenhagen measurement animation A wave packet reaches a detector, one detector record appears, and a probability notebook updates without a pulse traveling to the distant laboratory. prediction apparatus record one click probabilities update No physical message is drawn across space.
What moves: predictions become a detector record. In operational Copenhagen, "collapse" is usually an update rule, not a new controllable force.
01

Copenhagen interpretation

An operational family: use the theory, and be cautious about the picture behind it.

Quantum mechanics is a rule for assigning probabilities to possible measurement records. It need not provide a complete observer-independent variable λ that turns the distant outcomes into two local responses.

State as prediction tool No answers to unasked measurements Several versions

Changes

Declines pre-existing values for incompatible unperformed measurements and a fully mechanical account of the measurement cut.

Exists

Preparation procedures, measurement contexts, probabilities, and definite recorded outcomes.

Experiment

Standard quantum predictions. No separate Copenhagen deviation is expected.

Source: SEP, Copenhagen interpretation
Everett branching animation Local measurement interactions create multiple record branches. For the example equal-correlation setting pair A0 and B0, ordinary record-transfer paths later converge on matching comparison records inside overlapping light cones. ψ A:+ A:- B:+ B:- A0/B0 +/+ A0/B0 -/- one equal-correlation context; parity changes with the settings
What moves: local systems branch, then ordinary records travel to later comparison events. There is no instant cross-lab selection.
02

Many-Worlds

Everett's interpretation: keep the wave equation and give up one unique global outcome.

The universal wavefunction never collapses. Measurement is an ordinary local interaction that entangles observer, apparatus, and system into decohered branches containing every outcome.

Many outcomes Wave equation never breaks Locality disputed

Changes

The premise that each run produces exactly one absolute result pair for the entire universe.

Exists

One universal wavefunction and branch-relative records produced by decoherence.

Experiment

Designed to recover standard quantum statistics; no collapse deviations.

Source: SEP, Everettian quantum mechanics
Pilot-wave guidance animation A separate configuration-space panel contains the joint guiding law, while two actual particle positions move along definite paths in physical space below. configuration space (q1, q2) joint law, not a channel in 3-space physical-space positions particle 1: one actual path particle 2: one actual path
What moves: actual particle positions. The upper panel is a non-spatial configuration-space law, not a transmitted pulse.
03

Pilot-wave theory

de Broglie-Bohm mechanics: definite particles follow paths guided by a nonlocal wave.

Particles always have positions. A wavefunction guides their trajectories, but for entangled systems the velocity of one particle depends on the complete distant configuration.

Allows distant dependence One definite outcome No faster-than-light messages

Changes

The hidden guiding law is nonlocal: a distant setting can affect the joint guidance. Formally, parameter independence fails.

Exists

Actual particle configuration plus a guiding wavefunction, often with extra spacetime structure.

Experiment

Quantum equilibrium reproduces ordinary QM. Standard Bohmian mechanics adds no deviation.

Source: SEP, Bohmian mechanics
Objective collapse animation Two macroscopic wavefunction lobes coexist until stochastic localization noise suppresses one branch. here there stochastic localization is a new physical law GRW: discrete hits   |   CSL: continuous noise
What moves: a real stochastic localization field suppresses macroscopic superpositions. This is a modified theory, not only new language.
04

Objective collapse

GRW and CSL: make collapse a real, random, experimentally vulnerable process.

The Schrödinger equation is not exact. Rare spontaneous localizations, or continuous localization noise, force macroscopic systems toward one outcome while preserving microscopic interference for long enough.

Wave equation can break Still not a local Bell model Testable changes

Changes

Adds stochastic dynamics and one real outcome. Entangled collapse remains nonfactorizable.

Exists

A physical wavefunction plus collapse events, mass density, or a related primitive ontology.

Experiment

Predicts tiny excess decoherence, noise, or heating. Current tests constrain model parameters.

Source: SEP, collapse theories
Relational quantum mechanics animation Facts appear on relationships between physical systems. A later interaction produces a shared comparison record. Alice Bob pair compare fact relative to A fact relative to B shared fact exists at this later interaction
What moves: facts form at interactions. "Outcome" is a relation between systems before it is part of a shared comparison record.
05

Relational quantum mechanics

Facts are real, but always facts relative to another physical system.

There is no single observer-independent state assigning absolute properties everywhere. Interactions create events relative to the systems involved, and later interactions create new comparison facts.

Facts depend on the relationship No God's-eye state Observer means any physical system

Changes

The demand for one God's-eye joint state containing all distant outcomes as absolute facts.

Exists

Physical systems, interactions, and relational events. No mind is required.

Experiment

Uses standard quantum predictions; no established distinct Bell-test deviation.

Source: SEP, relational quantum mechanics
QBist probability update animation A probability assignment is shown inside an agent boundary. A local detector click changes the agent's probability bars. agent's expectations A experience p(+) p(-) The wavefunction is not drawn outside the agent as a physical field.
What moves: an agent's probability assignments after an experience. QBism does not propose a hidden signal between wings.
06

Quantum Bayesianism

QBism: the quantum state is an agent's probability assignment, not a distant physical substance.

Born probabilities are personal, normative expectations for the consequences of an agent's interventions. An outcome is that agent's experience, so Bell probabilities are not a mechanical common-cause model waiting to be factorized.

State as an agent's information Outcome as an agent's experience Not solipsism

Changes

The idea that ψ and Born probabilities describe agent-independent properties carried by the pair.

Exists

Agents, external systems, interventions, experiences, and coherence constraints on probabilities.

Experiment

Uses ordinary quantum probabilities; no standard QBist deviation.

Source: SEP, quantum Bayesianism
Consistent histories framework animation Two incompatible decoherent history frameworks alternate. A lock prevents propositions from both frameworks being combined into one story. framework α framework β Do not merge incompatible frameworks into one counterfactual ledger.
What moves: the framework used to ask a quantum question. Probabilities are valid within a decoherent family, not across an arbitrary mashup.
07

Consistent histories

Use one internally consistent quantum story at a time; do not mix incompatible stories.

Closed systems can be assigned probabilities over decoherent histories. But incompatible frameworks cannot be combined into one classical story containing answers to every counterfactual measurement question.

Do not combine incompatible stories Formal rule: one framework at a time Not a clean Bell loophole

Changes

Which collections of propositions may be combined into one probability space.

Exists

Quantum histories and framework-relative probabilistic descriptions of closed systems.

Experiment

Standard quantum predictions. Bell factorization still fails if one demands one local common-cause account.

Source: SEP, consistent histories
Superdeterministic common-cause animation A common region in the deep past sends forward causal links to the source variables and to both setting generators. common past λ Alice setting a Bob setting b ρ(λ | a,b) ≠ ρ(λ) through prior correlations
What moves: ordinary forward causal influences from a shared past. The key claim is statistical dependence, not a last-second conspiracy signal.
08

Superdeterminism

The detector choices and the pair can share causes in their common past.

Variables relevant to the source are correlated with the later measurement settings through common causes, global dynamics, or fundamental law. Bell's measurement-independence premise therefore does not apply.

Choices linked to the pair Shared causes in the past Tests depend on the model

Changes

ρ(λ|a,b)≠ρ(λ). The technical claim is correlation, not a semantic claim about all human agency.

Exists

Usually a deeper deterministic dynamics or common-past structure linking source and setting generators.

Experiment

A research program, not one theory. Concrete models may have testable residual structure.

Source: Hossenfelder and Palmer, Frontiers in Physics (2020)
Time-symmetric spacetime animation Time runs upward and space runs horizontally. Spacelike-separated future settings constrain a shared source in their past, while the source also connects forward to the later outcomes. future past space light-cone boundaries spacelike-separated settings setting a setting b A B λ shared source two-time constraint, not a controllable message to yesterday
What moves: in many models, boundary information runs both directions in time along spacetime paths. The settings remain side by side in the same future slice.
09

Time-symmetric theories

Retrocausal models: future detector settings can help constrain the pair's earlier state.

The experiment is treated as one spacetime boundary-value problem. Detector settings can constrain variables at the source through backward-in-time influence or an all-at-once consistency rule.

Future can constrain the past Choices and pair become linked Settings can still be genuine choices

Changes

The forward-only causal story behind measurement independence, often yielding ρ(λ|a,b)≠ρ(λ).

Exists

Two-time boundary conditions, time-symmetric fields, or globally constrained spacetime histories.

Experiment

Most proposals target standard QM; predictions beyond it depend on the concrete model.

Source: Wharton and Argaman, Rev. Mod. Phys. 92 (2020)
04 / FIELD GUIDE

Same evidence, different costs

No row gets every classical comfort for free.

This matrix is a map, not a verdict. "Disputed" means the classification changes with the precise definition of locality, fact, or ontology being used.

View Main idea Keeps a nearby-settings + shared-past explanation? One outcome for everyone? Does the wave equation always hold? Is the wavefunction a physical thing? Are settings independent of the pair? Can the future affect the past? Different experimental prediction?
Copenhagen Predictions and records No local common-cause model Yes, records Not across collapse Usually no claim Yes No No
Many-Worlds All outcomes branch Disputed No, branch-relative Yes Yes, universal Yes No No
Pilot-wave Particles plus a nonlocal guide No, the guiding law is nonlocal Yes Yes Yes Yes No No, in equilibrium
Objective collapse Real random collapse No Bell-local account Yes No Usually yes Yes Usually no Yes
Relational QM Facts exist between systems Reframes the global account Relative to systems Relational; no universal state No absolute ψ Yes Usually no No
Quantum Bayesianism An agent's probabilities No underlying causal model offered Agent experience Not an ontic law No Yes No ontic causal claim No
Consistent histories One framework at a time Does not restore Bell locality Within a framework Yes Usually yes Yes Usually no No
Superdeterminism Pair and settings share causes Can keep local causes Usually yes Model-dependent Model-dependent No No Model-dependent
Time-symmetric Past and future jointly constrain Can use local spacetime paths Usually yes Model-dependent Model-dependent Effectively linked Yes Model-dependent

"Purely local" means a complete shared-past account makes each wing respond only to its nearby setting. The wavefunction column asks whether ψ is itself an agent-independent physical thing, not whether an external world exists.

05 / THE VERDICT

What the experiments settle

With independent detector choices, Bell tests rule out a complete shared-past explanation in which each wing responds only to its nearby setting.

They do not choose one interpretation for us. Physics still needs an account of outcomes, causation, probability, and what the wavefunction represents.

Ruled out

The nearby-causes-only package

  • The detector choices are independent of the pair
  • The shared past contains everything needed to explain the outcomes
  • Each detector responds only to nearby causes
  • Those three claims cannot reproduce the observed correlations together
Still true

No faster-than-light messages

  • A remote setting does not change the local outcome distribution
  • That local distribution may itself be biased or unbiased
  • The correlation appears only after ordinary comparison
  • So "not Bell-local" does not mean "usable instant communication"
Still open

What kind of world explains it?

  • Nonlocal law, many outcomes, relational facts, or altered causation?
  • Is the wavefunction physical, relational, or epistemic?
  • Does exact unitarity ever fail?
  • Can a sharper model distinguish itself experimentally?

Bell did not make quantum mechanics vague. He showed exactly why nearby settings plus the shared past are not enough.

Nature keeps no faster-than-light telephone, but it refuses that Bell-local common-cause story.

Sources and further reading

The taxonomy follows the mathematical assumptions rather than the slogan "local realism." Interpretive placements are simplified for a field guide and explicitly marked where disputed.