What Measurement Formalism Does Not Settle

Quantum measurement formalism is an input–output calculus. Once a state and a measurement model are specified, it predicts outcome probabilities, conditional states, unread outputs, and sequential statistics.

That is already a powerful and experimentally successful structure. It is not, by itself, a microscopic detector theory, a derivation of the measurement model, an ontology of individual outcomes, or a proof of one interpretation of quantum mechanics.

This page makes those boundaries explicit. Its purpose is not to weaken ordinary measurement calculations, but to prevent a correct formula from being asked to answer a different kind of question.

What the Formalism Does Settle

For a finite-outcome instrument $\{\mathcal I_i\}$ and input state $\rho$, the operational rules determine

$$p(i) = \operatorname{Tr}\mathcal I_i(\rho),$$

the selected state

$$\rho_i = \frac{\mathcal I_i(\rho)}{p(i)}$$

when $p(i)>0$, and the unread state

$$\rho' = \sum_i\mathcal I_i(\rho).$$

For a later instrument $\{\mathcal J_j\}$, they also determine the ordered joint distribution

$$p(i,j) = \operatorname{Tr} \left[ \mathcal J_j \left( \mathcal I_i(\rho) \right) \right].$$

In the projective Lüders special case,

$$\mathcal I_a(\rho) = P_a\rho P_a.$$

These are precise claims. Given a valid model, they can be checked against frequencies, tomography, and subsequent measurements.

The formalism also imposes consistency conditions: effects must be positive and complete, instruments must be completely positive, and the unread map must preserve trace. These conditions rule out many proposed probability or update rules.

The Common Simplified Statement

Introductory presentations often say:

Measuring observable $A$ yields one of its eigenvalues and collapses the state into the corresponding eigenspace.

For ideal projective calculations, this is an efficient rule. It tells a reader which projectors to use and which conditional state enters later predictions.

The sentence becomes misleading when treated as a complete physical account. It silently assumes:

• that the laboratory device realizes the intended PVM;
• that the associated instrument is the Lüders instrument;
• that the outcome labels have been calibrated to the quoted eigenvalues;
• that “collapse” has an agreed physical or epistemic meaning;
• that the system–apparatus boundary has already been chosen;
• that detector inefficiency, amplification, and environmental records can be ignored.

A better operational statement is:

For an ideal projective measurement modeled by $\{P_a\}$ with the specified update instrument, the Born rule gives the probability of each recorded label and the conditional update used for later predictions.

That statement is less dramatic and more exact.

Formal Rule Versus Detector Mechanism

An ideal PVM $\{P_a\}$ or POVM $\{E_i\}$ is a mathematical model of the device’s outcome statistics. It does not specify how a real apparatus produces those statistics.

A detector description can involve several levels:

• Target quantity: the observable, parameter, or discrimination task the experiment is intended to probe.
• Probability model: PVM or POVM effects linking input states to reported outcomes.
• Backaction model: an instrument linking inputs to conditional quantum outputs.
• Microscopic dynamics: a Hamiltonian or open-system model for system, probe, amplifier, and environment.
• Classical response: thresholds, digitization, dead time, dark counts, filtering, and event selection.
• Calibration layer: trusted standards, parameter estimates, drift models, and uncertainty budgets.

The abstract formalism constrains each quantum layer after it is specified. It does not infer the correct layer from the name printed on the apparatus.

For example, a device advertised as a spin-$z$ detector need not realize the sharp effects $P_+$ and $P_-$. With symmetric label error $\epsilon$, its reported effects might instead be

$$\begin{aligned} E_+ &= (1-\epsilon)P_+ + \epsilon P_-,\\ E_- &= \epsilon P_+ + (1-\epsilon)P_-. \end{aligned}$$

The value of $\epsilon$ comes from detector modeling and calibration. The effects still do not determine the state update. Two instruments can share these response probabilities while disturbing the system differently.

A Unitary Premeasurement Model

A useful microscopic idealization begins with a system state

$$\lvert\psi\rangle = \sum_a c_a\lvert s_a\rangle$$

and an apparatus ready state $\lvert A_{\mathrm r}\rangle$. A unitary interaction correlates system alternatives with pointer states:

$$\left( \sum_a c_a\lvert s_a\rangle \right) \lvert A_{\mathrm r}\rangle \longmapsto \sum_a c_a \lvert s_a\rangle \lvert A_a\rangle.$$

For a basis input $\lvert s_a\rangle$, the pointer records $a$. By linearity, a superposed input produces an entangled state rather than one term chosen by the unitary equation.

If the pointer states are orthogonal, the apparatus reduced state is

$$\rho_A = \sum_a \lvert c_a\rvert^2 \lvert A_a\rangle\langle A_a\rvert.$$

This reduced state predicts a classical probability distribution for pointer measurements. Yet the composite state can remain

$$\lvert\Psi_{SA}\rangle = \sum_a c_a \lvert s_a\rangle \lvert A_a\rangle.$$

The unitary model explains correlation. By itself, it does not add a rule selecting one term as the unique global outcome.

Von Neumann Measurement Model develops the dynamics and idealizations. The historical framing belongs in Measurement Problem as Historical Problem.

The Measurement Problem

Several questions are often grouped under “the measurement problem”:

1. Why are outcomes represented by a particular stable set of macroscopic records rather than arbitrary superpositions?
2. How does a linear evolution of a superposed input relate to the experience or record of one definite result?
3. Why do outcome frequencies follow the Born weights?
4. What physical status should be assigned to the conditional state update?

Different research programs divide and answer these questions differently. The basic probability-and-update rules do not, merely by being written down, decide among those answers.

The Born rule answers:

Given $\rho$ and a measurement model, what is $p(i)$?

The selective update answers:

Given record $i$, what state predicts later records?

Neither formula alone answers:

Why did this run have this one record?

The third question is not obtained by algebraically rearranging the first two.

Proper and Improper Mixtures

The diagonal reduced state produced by entanglement can be operationally indistinguishable, for measurements on the subsystem alone, from a classical ensemble with the same density operator. The two preparation stories are nevertheless different.

A proper mixture represents classical uncertainty over alternatives in an ensemble. An improper mixture is the reduced state of a subsystem entangled with another system. For example,

$$\lvert\Psi\rangle = \frac{ \lvert0\rangle_S\lvert0\rangle_E + \lvert1\rangle_S\lvert1\rangle_E }{\sqrt2}$$

has reduced state

$$\rho_S = \frac12 \lvert0\rangle\langle0\rvert + \frac12 \lvert1\rangle\langle1\rvert.$$

The same $\rho_S$ could describe a proper half-and-half ensemble. No measurement on $S$ alone distinguishes the two. A joint measurement on $S\otimes E$ can distinguish the entangled pure state from the corresponding separable mixture.

Therefore diagonalization of a reduced state does not by itself prove that the global state has become one unknown but definite branch. Proper and Improper Mixtures is the canonical detailed treatment.

What Decoherence Explains

Environmental decoherence extends the premeasurement model. Suppose $\lvert\Psi_0\rangle\longmapsto\lvert\Psi\rangle$, with

$$\begin{aligned} \lvert\Psi_0\rangle &= \sum_a c_a \lvert s_a,A_a,E_0\rangle,\\ \lvert\Psi\rangle &= \sum_a c_a \lvert s_a,A_a,E_a\rangle. \end{aligned}$$

After tracing out the environment,

$$\rho_{SA} = \sum_{a,b} c_ac_b^* \langle E_b\rvert E_a\rangle \lvert s_a,A_a\rangle \langle s_b,A_b\rvert.$$

When

$$\langle E_b\rvert E_a\rangle \approx 0 \quad \text{for }a\neq b,$$

the interference terms are strongly suppressed in local observables. This explains several important features:

• why interference between macroscopically distinct records is difficult to observe;
• why reduced states become approximately diagonal in dynamically selected pointer structures;
• why records can be stable and redundantly copied into the environment;
• why classical stochastic descriptions become extraordinarily accurate for many macroscopic purposes.

These are dynamical achievements, not semantic relabelings.

What Decoherence Does Not Establish by Itself

Environmental decoherence under unitary dynamics does not, by itself:

• select one term of the global entangled state as the unique realized outcome;
• turn an improper mixture into a proper mixture;
• specify whether the state is ontic, epistemic, relational, or something else;
• derive the Born rule without additional assumptions or arguments;
• uniquely fix an exact preferred basis in every degenerate or approximate setting;
• choose a detector threshold or define which microscopic event counts as the recorded outcome.

Saying “decoherence is not collapse” does not mean decoherence is irrelevant. It means suppression of accessible interference and selection of stable records are not logically identical to a literal single-outcome postulate.

Start with Decoherence Preview, then continue to What Decoherence Does Not Solve.

Interpretations and Modified Theories

The same operational probabilities can be embedded in different conceptual frameworks. These frameworks can differ over the meaning of the state, the ontology, the scope of unitary dynamics, and the status of the update rule.

It is useful to distinguish broad kinds of proposal without pretending that labels alone settle their details:

• Operational or epistemic accounts treat the update primarily as a change in the state used for predictions, relative to a record or agent’s information.
• Everett or relative-state accounts retain universal unitary dynamics and explain definite-looking records through branch-relative structure.
• Hidden-variable theories supplement the wavefunction with additional variables or ontology that select actual configurations or events.
• Objective-collapse theories modify standard unitary dynamics with stochastic collapse processes.
• Other relational, perspectival, or reconstruction programs revise which states, facts, or principles are taken as fundamental.

There are substantial differences within every category. A textbook projection formula does not prove any one of them.

Some interpretations are designed to reproduce the same tested operational predictions in their intended domains. Modified-dynamics theories can make different predictions and are empirical theories, not merely alternative wording. Claims about experimental distinguishability must therefore identify the specific model and parameter regime.

Conditional States Are Not Signals

Measurement on one part of an entangled system can change the state conditioned on a remote outcome. That does not provide controllable faster-than-light communication.

Let $\rho_{AB}$ be bipartite, and let a local instrument on $A$ have Kraus operators $\{M_{i\alpha}\}$. Define the local unread channel and completeness operator by

$$\mathcal E_A(X) = \sum_{i,\alpha} M_{i\alpha}X M_{i\alpha}^\dagger,$$

and

$$C_A \equiv \sum_{i,\alpha} M_{i\alpha}^\dagger M_{i\alpha} = I_A.$$

If the outcome is ignored, Bob’s state becomes

$$\begin{aligned} \rho_B' &= \operatorname{Tr}_A \left[ (\mathcal E_A\otimes\operatorname{id}_B) (\rho_{AB}) \right]\\ &= \operatorname{Tr}_A \left[ \rho_{AB} (C_A\otimes I_B) \right]\\ &= \operatorname{Tr}_A\rho_{AB} = \rho_B. \end{aligned}$$

The second line uses cyclicity for operators acting within the subsystem being traced out, and the third uses completeness.

For a selected record $i$, the conditional state

$$\rho_{B\mid i} = \frac{ \operatorname{Tr}_A \left[ (\mathcal I_i\otimes\operatorname{id}_B) (\rho_{AB}) \right] }{ p(i) }$$

can differ from $\rho_B$. Bob can sort data into those conditional ensembles only after learning Alice’s record through an ordinary classical channel.

The update rule therefore captures changed conditional predictions without specifying a superluminal mechanical influence. Questions about the ontology of nonlocal correlations belong beyond the bare no-signaling calculation.

Probabilities, Frequencies, and Calibration

The formalism assigns a single-trial probability $p(i)$ once $\rho$ and $E_i$ are given. Under repeated independent trials with the same model, standard probability theory predicts concentration of relative frequencies around $p(i)$.

This does not make model validation automatic. Real data can violate ideal independent-and-identically-distributed assumptions because of:

• preparation drift;
• detector memory or dead time;
• unmodeled correlations;
• selection bias;
• changing backgrounds;
• incorrect Hilbert-space truncation;
• uncertainty in the effects themselves.

Agreement between observed frequencies and one predicted distribution does not uniquely identify the underlying instrument. Detector tomography requires trusted probe assumptions and uncertainty analysis; it cannot conjure a unique microscopic model from counts alone.

The Born rule supplies probabilities conditional on a quantum model. Statistics and calibration test whether that model is adequate for the data.

The System–Apparatus Boundary

The elementary formalism often treats the system quantum mechanically and the apparatus as a source of classical records. A larger description can include the apparatus, observer, or environment inside a quantum state.

The equations do not identify one unique place where that enlargement must stop. Different system boundaries can be useful for different tasks, provided overlapping descriptions make compatible operational predictions.

This flexibility is not permission to omit assumptions. A calculation should state:

• which degrees of freedom are in the system Hilbert space;
• which variables are treated as classical controls or records;
• which environment is traced out;
• which measurement model connects the quantum system to reported data.

Moving the boundary changes the state space and dynamical model. It should not be described as a passive change of notation.

A Claim-Classification Rule

Before making a measurement claim, identify its type:

• Formal theorem: follows from declared states, effects, instruments, and dynamics.
• Modeling assumption: ideal projectivity, Markovian noise, a particular coupling, or a chosen system boundary.
• Empirical claim: calibrated efficiency, observed decoherence rate, detector fidelity, or exclusion of a parameter range.
• Interpretive claim: statement about what the state, branches, hidden variables, or collapse mean.
• Speculative claim: proposal without established empirical support or a complete predictive model.

Claims can be valuable in every category. Trust comes from labeling the category and evidence correctly, not from making every statement sound like a theorem.

Evidence Labels gives the broader claim-strength policy.

Practical Rule for Calculations

For ordinary measurement work:

1. State the system, preparation, and Hilbert-space assumptions.
2. State the measurement data: PVM, POVM, or instrument.
3. Distinguish ideal parameters from calibrated detector parameters.
4. Compute probabilities before conditioning.
5. Normalize only branches whose outcomes are selected and have nonzero probability.
6. Use the unread channel when the record is unavailable.
7. Retain the instrument for sequential predictions.
8. Separate a dynamical apparatus model from the abstract effect model.
9. Label interpretation-dependent language.
10. Do not claim that decoherence, update, or diagonalization establishes more than the calculation shows.

This discipline leaves standard calculations concise while keeping their meaning controlled.

Canonical Boundaries

Continue according to the question being asked:

• Measurement in the Formalism for the operational input–output structure.
• State Update Rule for selective, unread, and partially observed branches.
• POVMs: First Encounter for effect-level outcome probabilities.
• Generalized Measurements Overview for instruments and Kraus operators.
• Von Neumann Measurement Model for system–apparatus dynamics.
• What Decoherence Does Not Solve for the detailed decoherence boundary.
• Measurement Problem as Historical Problem for historical development.
• What the Postulates Do Not Say for broader postulate-level limits.

Common Mistakes

• Treating the Born rule as a microscopic detector model.
• Treating a selective update as an explanation of why one outcome occurred.
• Assuming a POVM fixes the post-measurement state.
• Saying that a diagonal reduced state is automatically a proper classical mixture.
• Saying decoherence either solves everything or explains nothing.
• Presenting an interpretation as a theorem of the operational formalism.
• Presenting modified collapse dynamics as merely a change of words.
• Interpreting a conditional remote state as a controllable signal.
• Assuming that the apparatus label uniquely identifies its effects or instrument.
• Inferring a unique mechanism from agreement with one set of outcome frequencies.
• Moving the system–apparatus boundary without changing the model.
• Using “observation” as if the formalism required human consciousness; standard measurement models do not.

References

• J. von Neumann, Mathematical Foundations of Quantum Mechanics, Princeton University Press, 1955.
• G. Lüders, “Über die Zustandsänderung durch den Meßprozeß,” Annalen der Physik 8, 322–328, 1951.
• J. S. Bell, “Against ‘measurement’,” Physics World 3(8), 33–40, 1990.
• T. Maudlin, “Three measurement problems,” Topoi 14, 7–15, 1995.
• W. H. Zurek, “Decoherence, einselection, and the quantum origins of the classical,” Reviews of Modern Physics 75, 715–775, 2003.
• M. Schlosshauer, “Decoherence, the measurement problem, and interpretations of quantum mechanics,” Reviews of Modern Physics 76, 1267–1305, 2005.
• M. Schlosshauer, Decoherence and the Quantum-to-Classical Transition, 2nd ed., Springer, 2019.
• P. Busch, P. Lahti, J.-P. Pellonpää, and K. Ylinen, Quantum Measurement, Springer, 2016.
• H. M. Wiseman and G. J. Milburn, Quantum Measurement and Control, Cambridge University Press, 2010.

Exercises

1. Classify measurement claims

Classify each statement as a formal theorem, modeling assumption, empirical claim, or interpretive claim:

1. $\sum_iE_i=I$ implies $\sum_i p(i)=1$.
2. This photodetector has efficiency $0.91\pm0.02$ over the specified wavelength band.
3. The apparatus is accurately modeled by a Lüders instrument.
4. The state update represents a literal physical collapse.

Solution

1. This is a formal theorem once $p(i)=\operatorname{Tr}(\rho E_i)$ and $\operatorname{Tr}\rho=1$ are declared.
2. This is an empirical calibration claim and should be supported by data and an uncertainty analysis.
3. This is a modeling assumption unless independently established by process or instrument characterization.
4. This is an interpretive claim in standard quantum mechanics. A specific objective-collapse model would promote collapse to modified dynamics with its own empirical parameters.

2. Separate response probabilities from backaction

A qubit detector has effects

$$\begin{aligned} E_0 &= (1-\epsilon)P_0+\epsilon P_1,\\ E_1 &= \epsilon P_0+(1-\epsilon)P_1. \end{aligned}$$

Explain what these operators determine and give two kinds of additional information needed to predict a later measurement.

Solution

The effects determine the reported probabilities

$$p(i) = \operatorname{Tr}(\rho E_i)$$

for every input $\rho$. To predict a later measurement, one needs an instrument specifying the conditional output operation for each reported outcome. If there is evolution between measurements, one also needs the intervening channel or Hamiltonian and elapsed time. The error parameter and effects alone do not determine either backaction or later dynamics.

3. Derive decoherence of a two-branch record

Let

$$\lvert\Psi\rangle = c_0\lvert0,A_0,E_0\rangle + c_1\lvert1,A_1,E_1\rangle.$$

Trace out the environment and identify the factor multiplying the off-diagonal record term.

Solution

The total projector contains the cross term

$$c_0c_1^* \lvert0,A_0,E_0\rangle \langle1,A_1,E_1\rvert$$

and its adjoint. Tracing over $E$ gives

$$\begin{aligned} \rho_{SA} &= \lvert c_0\rvert^2 \lvert0,A_0\rangle\langle0,A_0\rvert\\ &\quad+ \lvert c_1\rvert^2 \lvert1,A_1\rangle\langle1,A_1\rvert\\ &\quad+ c_0c_1^* \langle E_1\rvert E_0\rangle \lvert0,A_0\rangle\langle1,A_1\rvert\\ &\quad+ c_1c_0^* \langle E_0\rvert E_1\rangle \lvert1,A_1\rangle\langle0,A_0\rvert. \end{aligned}$$

The decoherence factor is $\langle E_1\rvert E_0\rangle$. Nearly orthogonal environment states suppress the local off-diagonal record terms.

4. Distinguish proper and improper mixtures

Compare

$$\rho_{\mathrm{mix}} = \frac12 \lvert00\rangle\langle00\rvert + \frac12 \lvert11\rangle\langle11\rvert$$

with

$$\rho_{\mathrm{ent}} = \lvert\Phi^+\rangle\langle\Phi^+\rvert, \qquad \lvert\Phi^+\rangle = \frac{ \lvert00\rangle+\lvert11\rangle }{\sqrt2}.$$

Show that the first subsystem has the same reduced state in both cases, then give a joint observable that distinguishes the global states.

Solution

Tracing out the second subsystem gives

$$\rho_A = \frac12P_0+\frac12P_1 = \frac12I$$

for both global states. No measurement on $A$ alone distinguishes them.

The joint observable $X\otimes X$ does distinguish them:

$$\operatorname{Tr} \left[ \rho_{\mathrm{ent}} (X\otimes X) \right] = 1,$$

whereas

$$\operatorname{Tr} \left[ \rho_{\mathrm{mix}} (X\otimes X) \right] = 0.$$

The equal reduced density operator does not make the global preparation stories equivalent.

5. Prove unread local measurement cannot signal

Let $\{M_{i\alpha}\}$ be a complete local instrument on subsystem $A$:

$$\sum_{i,\alpha} M_{i\alpha}^\dagger M_{i\alpha} = I_A.$$

Show that ignoring the result leaves subsystem $B$ in its original reduced state.

Solution

The unread output on $AB$ is

$$\rho_{AB}' = \sum_{i,\alpha} (M_{i\alpha}\otimes I_B) \rho_{AB} (M_{i\alpha}^\dagger\otimes I_B).$$

Taking the partial trace and cycling the $A$-local operators inside it,

$$\begin{aligned} \rho_B' &= \operatorname{Tr}_A \left[ \rho_{AB} \sum_{i,\alpha} (M_{i\alpha}^\dagger M_{i\alpha}\otimes I_B) \right]\\ &= \operatorname{Tr}_A \left[ \rho_{AB} (I_A\otimes I_B) \right]\\ &= \operatorname{Tr}_A\rho_{AB} = \rho_B. \end{aligned}$$

Selected subensembles on $B$ can change, but identifying them requires the outcome record from $A$.

6. Identify the unresolved step in premeasurement

A unitary interaction maps $\lvert\Psi_{\mathrm{in}}\rangle$ to $\lvert\Psi_{\mathrm{out}}\rangle$, where

$$\begin{aligned} \lvert\Psi_{\mathrm{in}}\rangle &= \left( \alpha\lvert0\rangle + \beta\lvert1\rangle \right) \lvert A_{\mathrm r}\rangle,\\ \lvert\Psi_{\mathrm{out}}\rangle &= \alpha\lvert0,A_0\rangle + \beta\lvert1,A_1\rangle. \end{aligned}$$

What has this equation explained, and what definite-outcome question has it not answered?

Solution

The equation explains how a measurement interaction correlates each basis alternative with a corresponding pointer record. It also predicts, after reduction or decoherence, the local statistics of pointer observables.

Under unitary evolution alone, a superposed input produces the displayed entangled superposition. The equation has not selected either the $A_0$ term or the $A_1$ term as the unique global result. How definite records are understood requires an update postulate, modified dynamics, additional ontology, or an interpretation of branch-relative records.

7. Diagnose an inference from frequencies

An experiment observes outcome frequencies close to the predictions of effects $\{E_i\}$. Explain why this agreement does not uniquely determine the detector’s instrument or microscopic Hamiltonian.

Solution

Outcome frequencies constrain the probability functionals

$$p(i\mid\rho) = \operatorname{Tr}(\rho E_i)$$

for the tested preparations. Many instruments can have the same effects and therefore the same single-measurement frequencies while producing different conditional states. Different microscopic system–apparatus models can also induce the same effective instrument over the tested regime. Sequential data, trusted probes, dynamical tests, and model assumptions are needed to constrain those additional layers.

8. Repair overclaims

Rewrite each statement so that it says no more than the formalism supports:

1. “Decoherence proves that the wavefunction really collapses.”
2. “The apparatus measures $A$, so its effects are exactly the spectral projectors of $A$.”
3. “Alice’s measurement instantaneously changes Bob’s local statistics.”

Solution

1. Environmental decoherence suppresses locally accessible interference between dynamically recorded alternatives; whether this is accompanied by literal collapse depends on additional dynamics or interpretation.
2. The apparatus is intended to measure $A$; calibration is required to establish how closely its effects approximate the spectral projectors and what error model applies.
3. Conditioning on Alice’s recorded outcome can change the state assigned to Bob’s selected subensemble, but Alice’s unread local measurement leaves Bob’s reduced state and local statistics unchanged.

Summary

Measurement formalism precisely predicts probabilities and conditional quantum outputs once a state and instrument are specified. It does not by itself choose the correct detector model, derive calibration parameters, select one global outcome from a unitary superposition, determine the ontology of the state, or prove one interpretation.

Decoherence explains the dynamical suppression of local interference and the stability of classical-looking records, but it does not automatically convert an improper mixture into a proper one or select a unique result. Conditional remote states do not enable signaling, and matching frequencies do not uniquely identify backaction.

The practical standard is simple: distinguish theorem, model, empirical claim, and interpretation; then make each claim only as strong as its assumptions and evidence allow.