Bell's Theorem: Local Causality & Quantum Nonlocality

• Bell's Theorem is a fundamental framework demonstrating that quantum mechanics' predictions for entangled states defy reproduction by local hidden-variable theories, as exemplified by the CHSH inequality and Tsirelson’s bound.
• It distinguishes between deterministic and probabilistic formulations by examining locality, local causality, and factorization conditions, further analyzed through causal DAG representations.
• Experimental tests confirm quantum violations of Bell inequalities, driving advances in quantum information and spurring debates on interpretative issues like no-signalling versus nonlocality.

Searching arXiv for Bell’s theorem sources and related reviews. arxiv_search query="Bell's theorem review CHSH nonlocality local causality" max_results=10

Bell’s theorem is the class of results showing that the statistical predictions of quantum mechanics for separated measurements cannot be reproduced by theories that combine locality with an appropriate hidden-variable description. In the standard Bell–CHSH form, local hidden-variable models imply a bound $|S|\le 2$, while quantum mechanics predicts violations up to $2\sqrt2$ for suitable measurements on entangled states (Brunner et al., 2013). The literature also distinguishes Bell’s 1964 theorem, formulated from locality and determinism, from Bell’s 1976 theorem, formulated in terms of local causality; the two are logically equivalent in empirical content, but they are not identical in assumptions or interpretive force (Wiseman, 2014).

1. EPR origins, Einstein’s separation principle, and the completeness question

Bell’s theorem emerged from the Einstein–Podolsky–Rosen debate over whether the quantum-mechanical wave function $\psi$ provides a complete description of physical reality. EPR introduced what Einstein later called the separation principle: if two systems are spatially separated and no longer interact, then a measurement on one cannot instantaneously affect the real state of the other. For certain entangled states, EPR argued that one can predict with certainty the outcome of different non-commuting observables on one subsystem by choosing different measurements on the other. If one insists both that the predictable quantity corresponds to an element of reality and that quantum mechanics is complete, one is driven to the contradiction that the same physical state would have to correspond to different wave functions. EPR’s conclusion was therefore that quantum mechanics is incomplete (Boughn, 2016).

This background matters because Bell did not begin from an abstract probabilistic puzzle. He formalized the EPR tension between separability, definite properties, and quantum correlations. In later reconstructions, this historical point is used to distinguish two communities of interpretation: one that treats Bell chiefly as a theorem about determinism plus locality, and another that treats Bell chiefly as a theorem about local causality as a unified causal principle (Wiseman, 2014).

2. Formal structure: hidden variables, locality, and the CHSH inequality

In Bell’s 1964 deterministic formulation, measurement outcomes are fixed by hidden variables $\lambda$. For settings $a$ and $b$ at two wings, the outcomes are represented by functions

$A(a,\lambda)=\pm1,\qquad B(b,\lambda)=\pm1,$

with locality requiring that $A(a,\lambda)$ be independent of $b$ and $B(b,\lambda)$ be independent of $2\sqrt2$0. If $2\sqrt2$1 is distributed according to a normalized nonnegative density $2\sqrt2$2, $2\sqrt2$3, then the correlation function is

$2\sqrt2$4

For four settings $2\sqrt2$5, one defines

$2\sqrt2$6

Using only $2\sqrt2$7, one obtains the CHSH inequality

$2\sqrt2$8

This is the canonical Bell bound for deterministic local hidden-variable models (Boughn, 2016).

Bell’s later formulation replaced deterministic outcome functions with a probabilistic condition on “beables.” In the simplest bipartite case, local causality is expressed by the factorization

$2\sqrt2$9

where $\psi$0 is a complete specification of relevant past beables. This generalization does not assume determinism: the local response functions may be stochastic. The factorization is commonly decomposed into parameter independence and outcome independence, both of which follow from local causality but play different conceptual roles. Under measurement-setting independence, CHSH-type inequalities follow in the same way as in the deterministic case (Brown et al., 2014).

A complementary formulation recasts Bell’s theorem in the language of causal DAGs. There the relevant nodes are the settings $\psi$1, the hidden variable $\psi$2, and outcomes $\psi$3, with arrows $\psi$4, $\psi$5, $\psi$6, and $\psi$7, but no cross-arrows $\psi$8 or $\psi$9, and no arrows between the settings and $\lambda$0. The causal-Markov factorization then yields the same local hidden-variable structure and hence the CHSH bound (Gill, 2022).

3. Quantum violation: singlet correlations, Tsirelson’s bound, and no-inequality variants

For two spin-$\lambda$1 particles in the singlet state

$\lambda$2

quantum mechanics predicts the correlation

$\lambda$3

Choosing coplanar settings such as

$\lambda$4

one finds $\lambda$5, which exceeds the local bound $\lambda$6 and saturates the Tsirelson bound (Hance et al., 2022).

The standard CHSH violation is not the only form of Bell contradiction. The Greenberger–Horne–Zeilinger argument uses perfect correlations of a three-qubit entangled state and derives an “all-versus-nothing” contradiction without inequalities. A local hidden-variable assignment to the relevant $\lambda$7 and $\lambda$8 observables yields mutually inconsistent sign constraints, because multiplying the left-hand sides gives $\lambda$9 while multiplying the right-hand sides gives $a$0 (Ali, 2022).

A temporal analogue also exists. For a single spin-$a$1 with Hamiltonian

$a$2

one can define time-dependent observables $a$3 at three times and derive a GHZ-type contradiction from realism plus noninvasive measurability, which plays the role of temporal locality. In this formulation, Bell-type impossibility is not restricted to spacelike separated subsystems but extends to macrorealistic descriptions across time (Ali, 2022).

Some authors emphasize that the quantum violation can be understood as arising from superposition rather than from any explicit superluminal mechanism in the formalism. In that reading, the angle-dependent Bell correlations are built into the entangled state prepared at the source, and the theorem constrains hidden-variable completions rather than the unitary quantum predictions themselves (Boughn, 2016).

4. Locality, local causality, no-signalling, and the meaning of “nonlocality”

A major source of interpretive divergence is that “locality” does not mean the same thing in all Bell discussions. In the 1964 theorem, locality is the condition that the outcome on one wing does not depend on the distant setting. In the 1976 theorem, Bell instead used local causality, a stronger screening-off condition on conditional probabilities given a complete specification of beables in the joint past. The shift matters because violation of a Bell inequality shows that no locally causal completion exists, while the claim that ordinary quantum mechanics itself is nonlocal is a separate issue (Brown et al., 2014).

The distinction between Bell local causality and operational no-signalling is central. Standard quantum mechanics violates Bell inequalities yet preserves no-signalling: marginal statistics at one wing are independent of the distant setting. Several authors stress that this is weaker than Bell’s local causality. In this operational sense, quantum mechanics can remain compatible with relativistic causality even though it violates Bell-type factorizability (Zukowski, 2015).

The literature also separates parameter independence from outcome independence. Jarrett’s decomposition identifies parameter independence with Bell’s 1964 locality condition, while outcome independence expresses the further requirement that, given the hidden variables and settings, one outcome carries no extra information about the other. Operational quantum theory violates outcome independence but respects parameter independence; Bohmian mechanics does the reverse. This decomposition clarifies why different interpretations diagnose the Bell conflict differently (Wiseman, 2014).

Interpretive proposals then diverge. One line, associated with Bell’s own later worries, holds that factorizability may be only a consequence of local causality, not its final formulation, and that the tension with special relativity is not fully settled by the mathematics alone. Another line argues that a fully Lorentz-covariant version of quantum theory, free of action-at-a-distance, can be articulated in the Everett interpretation: collapse is absent, dynamics remain local, and violation of factorization arises from non-separability and multiple outcomes rather than spacelike causal influence (Brown et al., 2014).

5. Experimental tests, loopholes, and quantum-information significance

Bell tests were historically designed as empirical discriminants between quantum mechanics and local hidden-variable theories. In the simplest account, experiments on entangled particles repeatedly confirm violations of Bell inequalities, so any hidden-variable completion that reproduces the observed statistics must abandon locality, predetermined outcomes, or measurement independence (Maccone, 2012). More detailed reviews distinguish the major loopholes: detection inefficiency, locality or freedom-of-choice failures, coincidence-window effects, and memory or finite-statistics issues (Brunner et al., 2013).

The detection and coincidence loopholes admit modified bounds. Under an overall efficiency $a$4, one has

$a$5

or

$a$6

To beat $a$7, the review literature states that one needs $a$8 for detection or $a$9 for coincidence (Gill, 2012). In parallel, finite-sample analyses derive exponentially small tail probabilities for local-realist exceedance of the CHSH bound under randomized setting choices (Gill, 2012).

Bell’s theorem also became foundational for quantum information science. The review literature treats Bell nonlocality as underpinning device-independent randomness generation, device-independent quantum key distribution, self-testing, and the geometric analysis of correlation sets. In that geometric picture, one distinguishes the local polytope $b$0, the quantum set $b$1—convex but not a polytope—and the no-signalling polytope $b$2; linear programming characterizes $b$3 and $b$4, while semidefinite programming via the NPA hierarchy provides outer approximations to $b$5 (Brunner et al., 2013). In a more application-focused formulation, the violation of Bell’s theorem is described as “a necessary and sufficient criterion for generating a secure key for cryptography at two distant locations” (Hiesmayr, 2015).

Extensions to high-energy systems broaden the experimental domain. For neutral kaon–antikaon pairs produced in entangled antisymmetric states, one must incorporate decay, active strangeness measurements at selected proper times, and the possibility that one or both mesons decay before the measurement. The resulting Bell inequality acquires an extra term,

$b$6

reflecting the decayed–undecayed subensembles. In this setting, a striking claim is that the possibility of violating the modified Bell inequality depends on CP violation in $b$7–$b$8 mixing and decay (Hiesmayr, 2015).

6. Alternative diagnoses, reformulations, and ongoing disputes

A large part of the modern Bell literature is not about the algebra of CHSH itself but about which premise must be relinquished. One explicit proposal holds that Bell’s theorem establishes only

$b$9

On this view, locality and realism can be preserved if one rejects statistical independence, so that $A(a,\lambda)=\pm1,\qquad B(b,\lambda)=\pm1,$0. Superdeterministic, retrocausal, and “supermeasured” models are cited as examples of this route (Hance et al., 2022).

Another proposal isolates classical logic rather than measurement independence as the hidden premise. In a three-valued framework with truth values $A(a,\lambda)=\pm1,\qquad B(b,\lambda)=\pm1,$1, Henson rejects ontic definiteness while retaining a reformulated Einstein Locality and freedom of settings. In that construction, superluminal signalling is still forbidden, yet Bell inequalities can be violated because no underlying joint 0–1 valuation for all counterfactual propositions is assumed (Henson, 2011).

A different reformulation links Bell-type restrictions to Moore’s theorem from classical automata theory. In that account, no finite observational record suffices to bound the hidden degrees of freedom of an environment or to guarantee that recorded outcomes come from a single fixed subsystem. This is used to derive no-cloning, Kochen–Specker, and Bell analogues from limits on system identification rather than from specifically quantum postulates (Fields, 2012).

There are also direct critiques of the standard Bell argument. One line argues that Bell’s theorem is based on circular reasoning because it assumes linear additivity of expectation values for non-commuting observables and thereby tacitly builds in the CHSH bound $A(a,\lambda)=\pm1,\qquad B(b,\lambda)=\pm1,$2; on that reading, the correct bound for contextual hidden-variable theories is $A(a,\lambda)=\pm1,\qquad B(b,\lambda)=\pm1,$3 rather than $A(a,\lambda)=\pm1,\qquad B(b,\lambda)=\pm1,$4 (Christian, 2023). Another line accepts the standard proof but argues that finite-sample experimental verification may be affected by ignorance of the hidden-variable distribution, yielding a possible “statistics loophole” in the empirical inference from observed $A(a,\lambda)=\pm1,\qquad B(b,\lambda)=\pm1,$5 to the asymptotic Bell bound (Aiello, 2024). These proposals are best understood as disputes about the scope of Bell inferences, not as changes to the canonical CHSH derivation itself.

A further recurring controversy concerns whether Bell violations force “spooky action at a distance.” Several authors argue that this conclusion does not follow if one rejects counterfactual definiteness or hidden-variable completion and instead accepts complementarity or quantum incompleteness. On these readings, Bell’s theorem shows that no locally causal hidden-variable completion reproduces quantum predictions, but it does not compel the conclusion that the standard quantum formalism itself contains superluminal physical processes (Zukowski, 2015). This suggests that Bell’s theorem functions both as a no-go theorem for classical local realism and as a diagnostic tool for clarifying what different interpretations mean by locality, causation, reality, and completeness.