eduKate Learning Manual
Science | Edge Cases Science | Physical World
Understand → Observe → Explain → Test → Transfer → Go Deeper

The Leggett–Garg Inequality

How Measurements Across Time Can Challenge a Classical Story of Definite States

Quick Answer

The Leggett–Garg framework begins with two classical-style assumptions:

• Macrorealism per se: at any time, a macroscopic system is in one definite available state.
• Non-invasive measurability: it is possible, in principle, to determine that state without changing the system’s subsequent dynamics.

For a dichotomic observable Q(t) taking values ±1, define temporal correlations such as

C_ij = ⟨Q(t_i)Q(t_j)⟩.

A common three-time Leggett–Garg combination is

K = C₁₂ + C₂₃ − C₁₃ ≤ 1

under the relevant macrorealist assumptions and standard setup.

Quantum mechanics can predict values above the classical bound.

The difficult part is not only observing K > 1. It is showing that the sequential measurement procedure did not itself disturb the system in a way that a classical model could use to explain the violation. This is the clumsiness loophole.

Physical Review A — Leggett–Garg Macrorealism and Temporal Correlations →

Physical Review Letters (2024) — Leggett–Garg Violation With Ideal Negative Measurements in Neutron Interferometry →

What You Will Learn

• What macrorealism means in the Leggett–Garg framework.
• What non-invasive measurability assumes.
• How temporal correlations are constructed.
• Where the inequality comes from.
• Why quantum coherence can violate the bound.
• Why measurement backaction creates a loophole.
• How ideal negative measurements try to address that loophole.
• Why stationarity and fair-sampling assumptions can matter in variants.
• How decoherence restores more classical temporal correlations.
• How Leggett–Garg differs from Bell inequalities.
• Why violation does not mean “the future changed the past.”
• How 2026 work continues refining temporal nonclassicality tests and their assumptions.

Part 1 — The Classical Story Being Tested

Imagine a macroscopic object that can be in one of two states, Q = +1 or Q = −1.

A strong classical realist story says that at every moment the object really occupies one definite state whether or not anyone looks.

The second assumption says we could, at least in principle, discover that state without affecting what happens later.

These assumptions allow a joint classical probability description for values at several times.

Part 2 — Why a Bound Appears

For any definite assignment Q₁, Q₂, Q₃ ∈ {−1,+1}, consider

Q₁Q₂ + Q₂Q₃ − Q₁Q₃.

Checking all possibilities shows that this quantity cannot exceed 1.

Averaging over many trials preserves the bound and gives the Leggett–Garg inequality.

The mathematics is simple. The experimental assumptions needed to justify combining the measured correlations are the difficult part.

Part 3 — Quantum Prediction

A coherently evolving two-level quantum system can occupy superpositions rather than one classical Q state at every intermediate time.

Its temporal correlation functions can therefore produce K above the macrorealist bound.

The violation shows that the observed temporal statistics cannot be reproduced by the full set of assumptions used to derive the inequality.

the experiment falsifies an assumption package, not one isolated philosophical sentence by itself.

Part 4 — Sequential Measurements Are Dangerous

To estimate C₁₂, we may measure at t₁ and t₂.

To estimate C₂₃, we may measure at t₂ and t₃.

But a measurement at t₁ can physically disturb the system before t₂.

A classical hidden-variable model could then say: “the violation occurred because your first measurement kicked the system.”

This is the core experimental vulnerability.

Part 5 — The Clumsiness Loophole

The clumsiness loophole is the possibility that an apparently nonclassical violation is caused by invasive measurement rather than by failure of macrorealism itself.

Closing this loophole perfectly is extremely difficult because every real measurement couples to the system somehow.

The scientific strategy is therefore to design measurements that a macrorealist should accept as non-invasive, or to quantify and bound the disturbance independently.

Part 6 — Ideal Negative Measurement

Leggett and Garg proposed a clever method.

Suppose a detector is arranged to interact only if the system is in state +1.

If the detector does not click, a macrorealist can infer the system was in state −1 and argue that no physical interaction occurred.

By combining complementary runs, one estimates the state while minimizing a classical disturbance explanation.

Modern neutron-interferometry experiments have used ideal-negative-measurement strategies to obtain clear LGI violations.

Part 7 — Why “Negative” Does Not Mean Quantum-Mechanically No Backaction

From a quantum perspective, even a no-click outcome can update the state.

The point is subtler: the Leggett–Garg test is asking whether a macrorealist model can explain the data while maintaining its own claim that the negative-result measurement did not physically disturb the system.

The method therefore targets the assumptions of the classical competitor model rather than claiming measurement-free magic.

Part 8 — Decoherence Weakens Violations

Environmental dephasing destroys coherent phase relationships.

As coherence decreases, temporal correlations tend to become more compatible with classical macrorealist bounds.

This makes LGI experiments useful for probing the quantum-to-classical transition.

But loss of violation does not prove the system became literally classical in every respect. It means the chosen temporal witness no longer detects a violation under the measured conditions.

Part 9 — Leggett–Garg vs Bell

Leggett–Garg	Bell
Correlations across different times.	Correlations across separated subsystems/locations.
Tests macrorealism + non-invasive measurability assumptions.	Tests local hidden-variable assumptions.
Main loophole: measurement invasiveness/clumsiness.	Main loopholes historically include locality and detection.
One system can be measured sequentially.	Usually two or more spatially separated systems.

The phrase “Bell inequality in time” is useful intuition but not a complete equivalence.

Part 10 — Stationarity Variants

Some experimental variants avoid explicit sequential measurement by assuming stationarity or known dynamics.

These methods can reduce one loophole while introducing another assumption.

A 2026 Physical Review A study of Tsirelson-style temporal inequalities emphasized exactly this issue: if an alternative inequality relies on a dynamical assumption such as uniform precession, the assumption itself must be tested rather than silently imported.

Physical Review A (2026) — Assessing Dynamical Assumptions in Temporal Nonclassicality Tests →

Part 11 — 2025–2026: Temporal Nonclassicality Is Still an Active Frontier

Recent work continues exploring how strongly temporal inequalities can be violated and how measurement assumptions should be interpreted.

In late 2025, Physical Review Letters reported extreme LGI violations for systems evolving under superpositions of unitaries.

In 2026, new studies examined the role of internal nonlinear dynamics, measurement assumptions and information erasure in macrorealism tests.

The field’s direction is important educationally: stronger violation is not automatically stronger evidence unless the assumption and loophole structure is equally strong.

Physical Review Letters — Extreme Violations of Leggett–Garg Inequalities →

Part 12 — Why Violation Does Not Mean Retrocausality

Temporal correlations involve measurements at different times, which makes sensational “future changes past” stories tempting.

But an LGI violation does not by itself imply information travelled backward in time.

It shows that the measured temporal statistics are incompatible with the specified macrorealist/non-invasive assumptions.

No retrocausal signalling mechanism is required.

Part 13 — Why the Test Is About Models, Not Vocabulary

A theory can use the word “realist” in many different philosophical senses.

The Leggett–Garg inequality targets a specific operational model class.

The scientifically clean question is therefore:

Can a model with definite temporal values and suitably non-invasive measurability reproduce these correlations under the tested controls?

This avoids turning an experimental inequality into an undefined metaphysical slogan.

Failed Model → Better Model

Naive model	Why it fails	Better model
K > 1 proves “reality is false.”	The inequality tests a specific assumption package.	State macrorealism and measurement assumptions explicitly.
Sequential measurements reveal pre-existing states without consequence.	Measurements can disturb later dynamics.	Control or bound invasiveness.
A large violation automatically closes loopholes.	Magnitude and assumption quality are separate.	Audit clumsiness, stationarity and sampling.
Temporal violation implies future-to-past signalling.	Correlation does not create a signalling channel.	Separate nonclassical temporal statistics from causality claims.

How Do We Know?

• Prepare the same initial state many times.
• Choose a dichotomic observable Q = ±1.
• Measure C₁₂, C₂₃ and C₁₃ in carefully designed ensembles.
• Compute the relevant Leggett–Garg combination.
• Use ideal-negative or nondemolition measurement strategies where possible.
• Measure disturbance with auxiliary controls.
• Vary coherence/decoherence strength.
• Test stationarity/dynamical assumptions independently.
• Compare against explicit macrorealist models that include allowed measurement disturbance.

Observation vs Inference

• Observation: temporal correlation combinations can exceed standard macrorealist bounds.
• Measurement: violations survive increasingly careful controls for invasiveness in several platforms.
• Inference: the tested data are incompatible with the corresponding macrorealist/non-invasive model class.
• Model: quantum coherent temporal evolution.
• Boundary: the strength of interpretation depends on how successfully measurement disturbance and other auxiliary assumptions are controlled.

Common Misconceptions

Misconception	Better model
Leggett–Garg is exactly Bell’s theorem in time.	It is analogous but tests a different assumption structure.
Ideal negative measurement means quantum mechanically no state update.	It is designed to be non-invasive from the macrorealist competitor’s perspective.
No violation means the system is classical.	The chosen witness may simply be insensitive under those conditions.
Violation proves retrocausality.	It establishes nonclassical temporal correlations under the tested assumptions, not backward signalling.

Checkpoint Questions

1. What is macrorealism per se?
2. What is non-invasive measurability?
3. What does C_ij measure?
4. Why does a classical bound appear?
5. How can quantum mechanics violate it?
6. What is the clumsiness loophole?
7. What is an ideal negative measurement?
8. Why is Leggett–Garg not identical to Bell?
9. Why can decoherence remove a violation?
10. Why does a violation not imply retrocausal signalling?

Answer Key

Open after attempting the questions

1. The system has a definite available macroscopic value at every time.
2. That value can, in principle, be measured without changing later dynamics.
3. The average product of dichotomic values measured at two times.
4. Definite ±1 assignments constrain which correlation combinations are possible.
5. Coherent superposition/interference produce temporal statistics outside that classical set.
6. The possibility that measurement disturbance creates the apparent violation.
7. Inferring a state from a deliberately non-interacting detector outcome in a way intended to satisfy a macrorealist non-invasiveness standard.
8. Bell concerns spatial locality; LGI concerns temporal realism and measurement invasiveness.
9. Loss of coherence suppresses the interference terms needed for strong temporal nonclassicality.
10. The inequality tests correlations and model assumptions, not a signalling mechanism into the past.

Primary Science Bridge

• measuring something can change what happens next;
• correlations across time can test a model;
• fair tests must control the effect of the measuring tool;
• breaking a rule tells you at least one assumption behind the rule failed;
• a surprising correlation is not automatically a causal message.

Secondary and JC Bridge

Core idea	Higher-resolution route
Repeated measurement	Temporal correlation function
Classical assumption	Macrorealism
Measurement quality	Non-invasive measurability
Bound	Leggett–Garg inequality
Quantum effect	Coherent temporal interference
Loophole	Clumsiness/backaction control

Unfamiliar Transfer Challenge

A superconducting device violates a Leggett–Garg bound, but the first measurement is known to shift the device frequency slightly.

Can the violation be interpreted immediately as failure of macrorealism? No. Quantify the frequency shift, build a classical invasive-measurement model, use a negative-result or nondemolition control where possible, and show that the violation survives the allowed disturbance.

Deep Science Window — Inequality as a Model Gate

An inequality compresses a large model class into a numerical boundary. A violation is powerful because it eliminates every model inside that class at once—but only if the experiment truly satisfies the assumptions used to derive the boundary. The gate is mathematical; the hard work is validating entry conditions.

Deep Science Window — Measurement Is Part of the Dynamics

Leggett–Garg tests force us to include the measuring apparatus in the causal story. A data point is not automatically passive observation. When measurements are sequential, their physical backaction can become part of the system’s future state and therefore part of the model.

Evidence Boundaries

• LGI violation ≠ “reality is false.”
• LGI violation ≠ proof of retrocausality.
• Large violation ≠ loopholes automatically closed.
• Ideal negative measurement ≠ quantum mechanically zero state update in every description.
• No violation ≠ system definitively classical.
• Temporal Bell analogy ≠ identical assumptions to spatial Bell tests.

Manual Summary — KNOW → CONNECT → EXPLAIN → APPLY → CHECK

KNOW: macrorealism, non-invasive measurability, temporal correlation, Leggett–Garg bound, quantum violation, clumsiness loophole.

CONNECT: definite-state assumptions to a correlation inequality, quantum coherence to violation, and measurement backaction to loophole analysis.

EXPLAIN: how measurements across time can challenge a classical story of definite states.

APPLY: evaluate whether a new temporal-correlation experiment truly excludes a macrorealist model.

CHECK: audit invasiveness, stationarity, sampling and alternative classical disturbance models before interpreting the violation.

Research Sources and Further Reading

• Physical Review A — Leggett–Garg Macrorealism and Temporal Correlations
• Physical Review Letters — Violation Using Ideal Negative Measurements in Neutron Interferometry
• Nature Communications — Violation With Ideal Non-Invasive Measurements
• Physical Review A (2026) — Assessing Dynamical Assumptions in Temporal Nonclassicality Tests