Preparation Independence in Quantum Ontology

• Preparation Independence (PI) is a principle stating that the joint ontic state of independently prepared quantum systems factorizes into individual distributions.
• It underlies the Pusey–Barrett–Rudolph theorem by enforcing disjoint supports for composite quantum state distributions and challenging psi-epistemic models.
• Alternative frameworks like no-preparation-signalling relax factorization, exposing measure-theoretic nuances and spurring new research on quantum state reality.

Preparation Independence (PI) is a central structural assumption in foundational arguments concerning the reality of the quantum state, particularly within the ontological models framework and the Pusey–Barrett–Rudolph (PBR) theorem. It formalizes the notion that independently prepared quantum systems should correspond to independent underlying ontic variables. The mathematical and conceptual status of PI has direct implications for the onticity or epistemicity of the quantum state, the structure of ontological models, and the possibility of constructing $\psi$-epistemic theories compatible with all quantum predictions.

1. Formal Definition and Role in Ontological Models

Preparation Independence states that when two quantum systems are individually prepared in states $\rho_1$ and $\rho_2$, the probability distribution over the joint ontic state factorizes: $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$ Using an alternative notation, if $\overline{p} = (p_1, p_2)$ and $\overline{\lambda} = (\lambda_1, \lambda_2)$,

$$h(\lambda_1, \lambda_2 \mid p_1, p_2) = h(\lambda_1 \mid p_1)\; h(\lambda_2 \mid p_2).$$

This assumption is not implied by quantum mechanics itself but reflects a natural independence expectation whenever systems are prepared without causal connection or communication. In ontological models, such factorization precludes the existence of ontic-level correlations between independently prepared subsystems.

PI is the crucial auxiliary assumption in the PBR theorem, which uses the independence hypothesis to constrain the overlap structure of ontic state distributions for composite product preparations, ultimately deriving consequences for the onticity of the quantum state (Mansfield, 2014).

2. Measure-Theoretic Subtleties and Dualized Reality Criteria

The assessment of onticity versus epistemicity hinges on support overlap between distributions $\mu_\psi(\lambda)$ for various $\psi$. The original Harrigan–Spekkens criterion specifies that the quantum state is ontic if for every pair $\psi \neq \phi$, $\rho_1$0. In continuous spaces, technical subtleties involving sets of measure zero arise—namely, whether "almost everywhere disjoint" is sufficient for onticity.

To remedy these difficulties, a dualized reality criterion has been introduced. Here, an $\rho_1$1-valued property $\rho_1$2 is termed ontic precisely when $\rho_1$3 is a delta distribution for each $\rho_1$4, thereby sidestepping the ambiguities of null sets and focusing directly on the structure of conditional distributions. This reformulation streamlines analysis in settings with nontrivial measure-theoretic structure (Mansfield, 2014).

3. Analogy with Bell Locality and Conceptual Significance

PI is structurally analogous to Bell-locality (factorizability) in measurement scenarios. Bell locality requires for any joint measurement $\rho_1$5 on a bipartite system and ontic state $\rho_1$6 that

$\rho_1$7

Preparation Independence imposes a formally similar constraint, but on the preparation side: the ontic representation of independently chosen state preparations must factorize. Both constraints prohibit ontological-level correlations that are not attributable to local choices, but while Bell-locality pertains to measurement events, PI pertains to the initial preparation of states. This formal analogy clarifies why PI, like Bell locality, plays a pivotal role in no-go theorems investigating the completeness or ontology of quantum mechanics. Notably, rejecting PI is conceptually parallel to postulating nonlocal correlations in hidden variable models (Mansfield, 2014).

4. Weaker No-Preparation-Signalling Principle

A strictly weaker alternative to PI is the no-preparation-signalling assumption. Rather than requiring full factorization, no-preparation-signalling only demands that the marginal distribution of each system’s ontic state depends on its own preparation choice, not on that of the distant party: $\rho_1$8 This allows arbitrary global correlations between the pair $\rho_1$9, so long as no causal influence can be exerted by varying the preparation on one system to affect the local distribution of the other. This principle maintains compatibility with Bell and Kochen-Specker forms of locality and prohibits superluminal signalling, yet disrupts the independence logic central to PI-based arguments (Mansfield, 2014).

5. Explicit Counter-Example: Violation of PI without Signalling

A concrete counter-example reveals the logical gap between PI and no-preparation-signalling. Let $\rho_2$0 be the overlap region for distributions $\rho_2$1 and $\rho_2$2 of single-system states $\rho_2$3, and set

$\rho_2$4

Define the joint distribution $\rho_2$5 so that $\rho_2$6, $\rho_2$7, $\rho_2$8, and $\rho_2$9. The marginals $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$0 are independent of $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$1, satisfying no-preparation-signalling, but $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$2, manifestly violating PI.

This construction blocks the PBR argument at the crucial inference step: PI is needed to conclude nonzero probability of the joint ontic state lying in $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$3 under all product preparations. In this model, no such situation ever occurs, so the standard PBR contradiction does not arise, yet no-signalling is preserved (Mansfield, 2014).

6. Independence Loophole Closure and Developments Without PI

Recent analysis has demonstrated that the conventional necessity of PI in single-system $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$4-onticity derivations is overstated. Once composite systems prepared in product states are known to have distributions with disjoint supports—i.e., joint $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$5-onticity—the onticity of individual subsystems follows directly from the tensor-product structure of quantum mechanics. In particular, Gao establishes that if the joint ontic distributions for all product states are supported on disjoint sets, then each marginal distribution is uniquely supported and hence individually $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$6-ontic, irrespective of any residual correlations (Gao, 25 Jan 2026).

This removes the key auxiliary assumption from the original PBR theorem, closing the so-called "PIP loophole" for $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$7-epistemic models. Any attempt to retain $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$8-epistemicity must now fail to reproduce the PBR result even for the composite system, forgoing correctness for entangled measurements on product states. This sharpens the foundation for $$\mu(\lambda_1, \lambda_2 \mid \rho_1 \otimes \rho_2) = \mu(\lambda_1 \mid \rho_1)\; \mu(\lambda_2 \mid \rho_2).$$9-ontology: the reality of the quantum state as an objective property of each subsystem is secured purely by the joint onticity of product preparations and quantum tensor-factorization, without a need for the full strength of PI (Gao, 25 Jan 2026).

7. Open Questions and Research Directions

The possibility of extending PBR-type arguments under only no-preparation-signalling remains unresolved. The main obstacle is that allowing global ontic correlations—while preventing signalling—can preclude the logical step from product overlap regions to outcome probabilities forbidden by quantum predictions. It is thus necessary to develop new arguments or measurements capable of ruling out overlaps without relying on joint-factorizability.

A plausible implication is that future progress on this question may require a finer characterization of allowed ontological correlations and potential anti-overlap measurement scenarios capable of restoring the $\overline{p} = (p_1, p_2)$0-onticity conclusion under the weaker no-preparation-signalling constraint. Such work would further clarify the minimal structural ingredients required for any no-go theorem against $\overline{p} = (p_1, p_2)$1-epistemic models compatible with quantum phenomena (Mansfield, 2014).