Outcome Problem

The outcome problem asks what in a theory represents the individual result recorded in one run. It persists even after the apparatus basis has been identified and the Born probabilities are known. The measurement-problem derivation supplies the global entangled state; this page focuses on the logical step from that state to a particular record.

Helpful background. Measurement Problem derives the entangled premeasurement state.

Three different descriptions of a record

For orthogonal record branches $|b_0\rangle$ and $|b_1\rangle$, distinguish the coherent state, the incoherent ensemble, and a conditioned branch:

$$\begin{aligned} |\Psi\rangle&=\alpha|b_0\rangle+\beta|b_1\rangle,\\ \rho_{\rm mix}&=|\alpha|^2|b_0\rangle\langle b_0| +|\beta|^2|b_1\rangle\langle b_1|,\\ \rho_{0}&=|b_0\rangle\langle b_0|. \end{aligned}$$

The first retains relative-phase coherence. The second describes statistics after coherence is discarded or after a stochastic preparation of branches. The third uses information that outcome zero was obtained. Passing from the first to the apparatus marginal is a mathematical operation; conditioning on a result uses an outcome event. Neither operation alone supplies a physical account of why that one event occurred.

An outcome is not required to be certain in advance. A stochastic theory can specify an event space and a law selecting one event without predicting which one. The issue is whether those individual events are part of the theory’s physical description, not whether the observer can know the future.

Why a diagonal matrix is insufficient

Take $\rho_A=I/2$ for a pointer idealized as a qubit. It can be written as an equal mixture of $|0\rangle,|1\rangle$, or an equal mixture of $|+\rangle,|-\rangle$. The density matrix alone therefore does not privilege one decomposition as the actually existing but unknown pointer value. Apparatus dynamics and its record observables matter.

The ambiguity is not restricted to equal eigenvalues. Applying the nonunique-ensemble construction to a pointer with probabilities $3/4$ and $1/4$ gives

$$\rho_A=\frac34|0\rangle\langle0|+\frac14|1\rangle\langle1| =\frac12|\chi_+\rangle\langle\chi_+|+\frac12|\chi_-\rangle\langle\chi_-|,$$

where

$$|\chi_\pm\rangle=\frac{\sqrt3}{2}|0\rangle\pm\frac12|1\rangle.$$

The cross terms cancel, while $\langle\chi_+|\chi_-\rangle=1/2$: the second ensemble consists of nonorthogonal states. The spectral basis is unique up to phases, but the preparation ensemble is not. No pointer-only measurement can determine which of these preparation procedures was used when its label is unavailable, since every effect $F$ has the same probability $\operatorname{Tr}(F\rho_A)$.

If $Z_A$ is the calibrated record observable, the $\chi_\pm$ vectors are superpositions of its record values, not additional definite $Z_A$ records. Thus ensemble nonuniqueness does not make every decomposition an equally good record basis; it shows why the density matrix alone cannot identify an actual but unknown preparation. Even the orthogonal spectral decomposition does not select one outcome of an entangled measurement. An interpretation can add a rule assigning definite properties using that structure; this is a further rule to assess, not the partial-trace identity itself.

How proposed accounts answer the question

A collapse law selects a stochastic state or matter configuration and must identify how it yields robust records. A pilot-wave account assigns an actual configuration, so a pointer can occupy one branch even when the wavefunction contains others. An Everettian account treats mutually decohered records as relative to branches and must explain the experience and probability of branch-relative outcomes. A relational account must say relative to which interacting system an event is definite and how later comparisons agree.

These claims should be evaluated using the same operational experiment. Merely replacing “outcome” with “information” does not specify the record-bearing degrees of freedom or the rule for their use. Conversely, treating every account as if it promised a unique global branch would misstate branch-relative proposals.

Exercise: conditional certainty

An ideal projective measurement gives outcome zero with probability $p$ and output state $|0\rangle$ conditional on that result. A second immediate measurement in the same basis returns zero with certainty. Does repeatability show that the first result had probability one?

Solution

No. The first probability is $p(0)=p$; repeatability states $p(0_{\rm second}\mid0_{\rm first})=1$. The unconditional probability of two zero records is still $p$. Conditional certainty after an event cannot be substituted for a premeasurement probability or a physical explanation of the event’s occurrence.

References

• Maudlin, Tim, “Three Measurement Problems,” Topoi 14, 7–15 (1995), doi:10.1007/BF00763473.
• Schlosshauer, Maximilian, Decoherence and the Quantum-to-Classical Transition, Springer (2007), doi:10.1007/978-3-540-35775-9 — reduced states, records and interpretive limits.