Frauchiger–Renner Argument

The Frauchiger–Renner argument combines quantum predictions with rules for transferring certainty between agents in an extended Wigner’s friend experiment. Its lesson is conditional: a specified collection of universal quantum reasoning, consistency rules, and single-outcome claims cannot all be applied as proposed. The algebra is precise; identifying which inferential rule an interpretation accepts requires separate analysis.

Two laboratories and two interference measurements

A first friend measures a quantum coin with probabilities $1/3$ for $h$ and $2/3$ for $t$. On $h$, the friend prepares a spin in $\lvert\downarrow\rangle$; on $t$, in $(\lvert\downarrow\rangle+\lvert\uparrow\rangle)/\sqrt2$. A second friend measures the spin. Include the records in orthonormal laboratory states $\lvert H\rangle,\lvert T\rangle$ and $\lvert D\rangle,\lvert U\rangle$. Before the external measurements, unitary evolution gives

$$\lvert\Psi\rangle=\frac{\lvert H,D\rangle+\lvert T,D\rangle +\lvert T,U\rangle}{\sqrt3}.$$

The external observers measure interference bases, with

$$\lvert\overline{\rm ok}\rangle= \frac{\lvert H\rangle-\lvert T\rangle}{\sqrt2}, \qquad \lvert{\rm ok}\rangle= \frac{\lvert D\rangle-\lvert U\rangle}{\sqrt2}.$$

The plus combinations are called $\overline{\rm fail}$ and $\mathrm{fail}$. The joint amplitude is $(1-1+1)/\sqrt{12}$, hence

$$p(\overline{\rm ok},{\rm ok})=\frac1{12}.$$

This is a prediction for an ideal coherent experiment, including the records. It is not a calculation in which both friends have already caused irreversible objective collapse.

The chain of certainty

Three probability-one inferences motivate the paradox:

1. Conditioning on $\overline{\rm ok}$ cancels the $\lvert D\rangle$ terms, leaving $\lvert U\rangle$.
2. A record $U$ implies the preparation was $T$, because there is no $\lvert H,U\rangle$ term.
3. Conditioning on $T$ before an interfering measurement leaves the second laboratory in $(\lvert D\rangle+\lvert U\rangle)/\sqrt2$, which predicts $\mathrm{fail}$.

If these agent-indexed statements are freely chained, $\overline{\rm ok}$ seems to imply $\mathrm{fail}$, contradicting the nonzero joint event. But each conditioning belongs to a particular description and measurement context. Inserting additional record measurements or external copies changes which coherences survive. The calculation therefore cannot be replaced by an ordinary joint distribution over all record and interference observables.

The published argument names assumptions Q, C, and S: quantum certainty assignments, consistency of an agent adopting another agent’s certainty, and exclusion of simultaneous certainty in incompatible single outcomes. Their exact operational use matters more than the shorthand “quantum mechanics, logic, and one world.”

Interpreting the obstruction

The global Born calculation is internally consistent. The disputed step is how agents can combine statements derived with different placements of the quantum/classical cut and different conditioning histories. del Rio and Renner’s reply and the discussion by Lazarovici and Hubert illustrate distinct assessments of these inference rules. One should not infer that every interpretation with definite outcomes is mathematically inconsistent, or that the experiment uniquely selects many worlds.

For a proposed response, specify the state used at each time, which records remain coherent, the intervention actually performed, and the rule permitting an agent to import another’s prediction. Objective collapse changes the dynamics; Everett changes the meaning of a unique global outcome; other accounts restrict the transfer of certainty or insist on an explicit total-system model.

Exercise

Project $\lvert\Psi\rangle$ onto $\lvert\overline{\rm ok}\rangle$ and normalize the surviving second-laboratory state. What is $p({\rm ok}\mid\overline{\rm ok})$?

Solution

The unnormalized state is $-\lvert U\rangle/\sqrt6$, so $p(\overline{\rm ok})=1/6$ and the normalized state is $\lvert U\rangle$ up to phase. Its $\mathrm{ok}$ probability is $1/2$, giving the joint probability $1/12$. Conditioning directly on the actual external outcome avoids the incompatible certainty chain.

References

• del Rio, L., and Renner, R., “Reply to: Quantum mechanical rules for observed observers and the consistency of quantum theory,” Nature Communications 15, 3024 (2024), doi:10.1038/s41467-024-47172-0 — clarification of the theorem’s intended scope.
• Frauchiger, D., and Renner, R., “Quantum theory cannot consistently describe the use of itself,” Nature Communications 9, 3711 (2018), doi:10.1038/s41467-018-05739-8 — original protocol and assumptions.
• Lazarovici, D., and Hubert, M., “How Quantum Mechanics can consistently describe the use of itself,” Scientific Reports 9, 470 (2019), doi:10.1038/s41598-018-37535-1 — an explicit counteranalysis.