Operational vs. Ontological Claims

Operational predictions and ontological explanations

An operational claim is framed in terms of laboratory-accessible procedures and their statistical relations:

• how a system is prepared;
• which measurement setting is selected;
• which outcome is recorded;
• what probabilities or frequencies are expected.

For a preparation procedure $P$, a measurement procedure $M$, and an outcome $k$, the basic operational object is

Read this as: given that we implement preparation $P$ and measurement $M$, what is the probability of record $k$?

An ontological claim, by contrast, says something about the physical constitution of the situation: what systems are made of, what state they are really in, and what processes produce the observed records. It addresses questions such as:

• Is the quantum state an objective physical field, or a representation of information?
• Does a particle have a definite position between measurements?
• Does a physical collapse occur, and if so, by what law?
• Are all measurement-result branches physically real?
• Is there a more complete state that fixes an individual outcome?

The contrast is not “empirical science” versus “empty metaphysics.” Operational statements can be precise, strongly testable, and technologically indispensable. Ontological statements can also be scientifically constrained. The distinction is instead about what the claim is about:

A useful caution: operational does not automatically mean “epistemic,” and ontological does not automatically mean “deterministic.”

• An operational formulation can simply refrain from saying what the quantum state represents.
• An epistemic interpretation goes further: it proposes that the state represents knowledge, belief, or information.
• An ontological model may be stochastic, not deterministic.
• A deterministic model is ontological only if it claims its determining variables describe physical reality rather than merely encode a prediction algorithm.

The formal distinction will be used repeatedly when we reach Bell, contextuality, hidden variables, Bohmian mechanics, and Everettian quantum mechanics.

What quantum theory says operationally

Consider a familiar prepare-and-measure experiment. A source prepares a qubit in

and a receiver measures in the $Z$ basis. The operational quantum prediction is

This means that, in many suitably controlled repetitions of that preparation-and-measurement procedure, the relative frequencies of the two detector records are expected to approach one half each.

For a general preparation represented by density operator ${\rho }_P$, and a measurement $M$ represented by POVM effects $\{E_k^M\}$, the operational prediction is the Born rule:

This expression is already rich. It permits quantitative calculations, communication-system design, tomography, error estimation, and experimental tests. But considered strictly as an operational rule, it does not by itself answer questions such as:

• Which physical property of the individual qubit makes one detector fire on a given trial?
• Is ${\rho }_P$ a physical state of the qubit, an ensemble description, or an agent-relative state assignment?
• Does the measurement interaction physically create the outcome, reveal a pre-existing feature, select one branch, or trigger a real collapse?

Those are added interpretive or ontological commitments.

Operational states are equivalence classes of procedures

An operational “state” need not refer to an invisible object directly. It can be understood as an equivalence class of preparation procedures. Two procedures are operationally equivalent when no available measurement distinguishes them:

For example, a polarization qubit produced by one optical circuit and the same density operator prepared by another physical implementation count as the same operational state if every permitted measurement yields the same outcome statistics.

This is an important move: the operational formalism deliberately compresses distinct physical implementations into the same predictive object whenever experiments cannot distinguish them at the relevant level. It thereby specifies what can be predicted from procedures, not necessarily the microscopic mechanism by which those procedures work.

That economy is useful, but it leaves open the ontological question. The two preparation devices may be physically identical in the relevant respects—or they may reach the same operational statistics through very different underlying mechanisms. The operational framework alone does not decide.

Adding an ontology: the hidden layer is a hypothesis

An ontological model introduces a space $\Lambda$ of possible complete physical states. Write a particular underlying state as

The preparation $P$ is represented by a distribution over these states,

and the measurement is represented by response probabilities,

The operational probabilities must be recovered by averaging over the possible ontic states:

For a discrete ontic state space, the integral becomes a sum.

This equation is not yet a specific interpretation of quantum mechanics. It is a framework for stating what an interpretation would have to do. Its components have different jobs:

• $\mu (\lambda \mid P)$ says what physical states are compatible with a preparation.
• $\xi (k\mid M,\lambda )$ says how those physical states, together with the measurement context, produce outcome probabilities.
• The integral ensures agreement with the empirically tested operational predictions.

The operational theory supplies the left-hand side. The ontological model supplies a candidate explanation of why that left-hand side has the value it does.

The diagram should not be read as proving that a hidden state exists. It shows the logical structure of an ontological-model proposal: it introduces additional variables and mechanisms, then must reproduce the observed distribution.

Determinism is one option inside an ontological model

The response rule can be genuinely probabilistic:

In that case, even a complete ontic state does not fix the result. The ontology is stochastic.

Alternatively, the model is outcome-deterministic if the complete physical state and measurement setting fix one outcome:

Here $f(M,\lambda )$ is the outcome selected by the physical state and measurement context. The apparent randomness in the operational distribution can then arise because the preparation leaves $\lambda$ unknown or distributed across runs.

This distinction is central to the course:

Operationally probabilistic predictions do not, by themselves, establish ontological indeterminism.

The same observed distribution can be compatible with a stochastic ontological account, a deterministic completion, or a view that declines to posit an underlying state at all. Later no-go results will show that proposed completions cannot satisfy every attractive constraint simultaneously—especially locality and noncontextuality. But that is a much more specific conclusion than “probabilistic data rule out determinism.”

One experiment, several incompatible stories

Return to the $|+⟩$ preparation followed by a $Z$-basis measurement. The operational content is straightforward:

Suppose repeated trials agree with this prediction within statistical uncertainty. That supports the operational claim. But consider several further claims one might hear:

The first two claims do not settle among the remaining ones.

For example, a deterministic completion might say that individual outcomes are fixed by a more complete state, while the quantum state summarizes a distribution over those states. A spontaneous-collapse theory might say that the outcome is objectively selected but fundamentally stochastic. An Everettian account might retain deterministic universal unitary dynamics while treating measurement as branching into decohered relative records. A deliberately operational interpretation may use the probability rule without claiming that the theory provides a microscopic description of what occurs between preparation and record.

These are not verbal variants of one theory. They offer substantially different answers to the ontological questions. Yet a simple prepare-and-measure data set need not discriminate among them.

The same discipline applies to locality

A particularly consequential example will arise with Bell experiments. The operational no-signalling condition is

It says that Alice’s observed marginal outcome distribution, conditioned on her setting $x$, does not depend on Bob’s freely varied setting $y$.

This is an operational statement about accessible probability distributions. It does not by itself entail the ontological statement

“There is no superluminal causal influence in the underlying physical world.”

Nor does an observed Bell violation, by itself, say that “reality is nonlocal” without qualification. Bell’s theorem instead rules out a package of assumptions about an underlying model. We will make that package explicit later: locality conditions, measurement independence, and the form of a common-cause explanation. Precision here prevents a great deal of foundational overstatement.

A caution about “what happens”

The preceding distinction can sound clean until one encounters ordinary physics language: “an electron emitted a photon,” “a gas thermalized,” or “a particle followed a trajectory.” Are these operational abbreviations for predicted records, or literal accounts of physical events?

The following seminar segment presents a forceful version of this concern. Its “category problem” is a philosophical diagnosis offered by the speaker, rather than an established theorem. Watch it to sharpen the distinction, while keeping in mind that realist interpretations reject its most restrictive instrumentalist conclusion.

The right conclusion is not that physicists must stop speaking of unobserved processes. Rather, one should state which level is intended. For instance:

• “Quantum mechanics predicts the observed spectral line intensity” is operational.
• “An electron physically makes a transition and emits a photon at a definite time” is ontological, unless explicitly offered only as shorthand.
• “A quantum field is the physical entity whose excitations account for the records” is an ontological proposal.
• “The state update is an agent’s rational update after obtaining a record” is epistemic or operational.

Re-reading the measurement problem with this distinction

The previous lesson can now be organized more carefully.

For a system $S$ and apparatus $A$, unitary premeasurement gives

Several different claims are embedded in discussions of this equation.

1. Operational claim:
   A measurement procedure has distinguishable pointer records $A_0$ and $A_1$, and repeated runs yield observed frequencies governed by the Born rule.

2. Formal dynamical claim:
   If the system and apparatus are treated as closed and evolve under the specified Hamiltonian, unitary evolution uniquely yields the displayed entangled state.

3. Ontological completeness claim:
   The quantum state in that equation is the complete physical state of the world.

4. Ontological outcome claim:
   Exactly one observer-independent outcome occurs in each run.

5. Interpretive explanatory claim:
   Collapse, hidden variables, branches, or some other physical structure explains the relation between the formal state and observed records.

The measurement problem arises because one cannot simply slide from claim 2 to claim 4. Unitary dynamics gives a unique state, but whether that state is a complete description, whether it contains one actual outcome, and what a record physically consists in are ontological questions.

Conversely, the plain fact that laboratory observers report definite records does not automatically select an ontology. It is operational input that every viable interpretation must recover. The competing interpretations differ in their explanation of that input.

A reliable habit is therefore to append a silent qualifier whenever encountering a foundational assertion:

• “Quantum mechanics is probabilistic” — at what level? Outcome statistics, state dynamics, or underlying events?
• “The wavefunction collapses” — as a calculational update or as a physical process?
• “The particle has no position before measurement” — as an operational restriction, or as an ontological denial?
• “Bell rules out locality” — which precise causal-model assumptions are included?

Key takeaways

• Operational claims concern procedures, records, and probabilities such as

  $$p(k\mid M,P).$$

• Ontological claims propose what physical systems are, what their complete states are, and what mechanisms produce observed records.

• The Born rule is operationally powerful:

  $$p(k\mid M,P)=\mathrm{Tr}\left(E_k^M{\rho }_P\right),$$

  but it does not on its own determine what ${\rho }_P$ represents physically.

• An ontological model introduces a complete state $\lambda$, preparation distributions $\mu (\lambda \mid P)$, and response functions $\xi (k\mid M,\lambda )$, while reproducing operational statistics.

• Operational randomness does not alone prove that underlying reality is indeterministic. A model may be stochastic, deterministic, or noncommittal about underlying physical states.

• Claims about no-signalling, outcome frequencies, and detector records should not be conflated with claims about causal influence, pre-existing properties, or the ultimate ontology.

Next, we will use this distinction to formulate the measurement problem as a precise incompatibility among three claims: completeness of the quantum state, universal unitary dynamics, and single definite outcomes.