Observing light changes an experiment when the observation introduces a physical interaction that changes the light, its correlations or the set of retained outcomes. The relevant question is which operation was performed and how its consequences enter the prediction.
Make the operation explicit
An absorbing detector, a polarization filter and a weakly coupled monitor perform different operations. They should not be grouped under an unexplained instruction to observe. A detector may consume the photon it registers; a filter may transmit only one polarization; a monitor may correlate the light with another system.
Reading an already stored display is another operation again. If it does not feed back into the optical apparatus, it does not change the earlier preparation or detection distribution. The physical chain can include a human reader, but the light responds to interactions, not to an unspecified state of awareness.
A simple arrangement of ideal polarizers makes the difference between an intervention and a passive description especially clear.
Three polarizers and a complete account
Start with light already polarized horizontally, with incident intensity normalized to one. Send it directly into an ideal vertical polarizer. The transmitted intensity is zero because the horizontal and vertical polarization vectors are orthogonal.
Now insert a diagonal polarizer at 45 degrees before the vertical one. The diagonal element transmits cos²45° = 1/2 of the incident intensity. The transmitted beam is diagonal. The vertical element then transmits half of that remaining intensity. The final output is 1/4 of the original.
The added polarizer has increased the transmission through the entire arrangement from zero to one quarter. It has not created energy. Half the input is removed at the diagonal element, another quarter at the vertical element and one quarter emerges. The three fractions sum to one.
These are ideal polarizers and a horizontally polarized input. If the initial light were unpolarized and a first horizontal polarizer were included in the accounting, that first element would introduce another factor of one half. Always state where the normalization begins.
Conditioning changes the denominator
For single photons prepared horizontally, the same ideal probabilities apply to repeated trials. Half pass the diagonal filter. Among those transmitted photons, half pass the vertical filter. The conditional probability for the second transmission is one half; the probability for both transmissions from the original input is one quarter.
Reporting only successful transmissions hides the rejected trials. This can be appropriate for preparing a selected output state, but it is not a complete account of the experiment's efficiency. Conditional and unconditional probabilities answer different questions.
The calculation also explains why the order matters. A vertical polarizer placed first removes every horizontally prepared photon in this ideal arrangement. Adding a diagonal polarizer afterwards cannot restore photons that have already been absorbed. Reordering physical operations is not merely reordering the sentences used to describe them.
Disturbance is more than a mechanical kick
A which-path interaction can change interference by correlating alternatives with distinguishable states of another system. Its consequence need not be explained as a large classical deflection. What matters for the accessible interference is whether the measurement arrangement still permits the relevant alternatives to combine coherently.
Conversely, attenuation alone does not always reduce normalized fringe visibility. Equal attenuation in two arms can reduce the number of detections while leaving the ideal normalized pattern unchanged. Unequal attenuation can reduce visibility through beam imbalance even when the surviving fields remain fully coherent.
The previous essay on what an interference pattern reveals separates those possibilities. A measured change in contrast therefore needs more diagnosis than the statement that someone looked.
Describe the record and the interaction
In the book's treatment of records and decoherence, an interaction correlates the system with an apparatus and environment. Tracing over unobserved degrees of freedom affects the accessible description. That calculation must be distinguished from a complete account of why a particular outcome is experienced in an individual trial.
For SMM/DTT, the useful task is to specify a preparation, an interaction, a completion rule and a readout, then recover the observed frequencies of every retained outcome. The polarizer calculation supplies a modest benchmark: a proposed model must preserve both the selected-state prediction and the full accounting of rejected inputs.
This approach makes an observation claim inspectable. Identify what was added to the apparatus, which alternatives became distinguishable, which trials were discarded and what probability changed. Each item can be connected to a calculation or an experimental control.
Sources and further reading
- Feynman, Leighton and Sands, Polarization and More two-state systems. The normalized loss ledger above applies the ideal projection rule.
- Mohamed Haj Yousef, Records and decoherence and What would test an interpretation?.