A camera records an empty room while everyone is elsewhere. Later, someone watches the recording. At what point did observation occur: when light reached the sensor, when the data were saved, or when a person understood the image?
The question becomes easier when we stop asking one word to perform every task. Detecting, recording and interpreting are connected activities, but an account can distinguish them.
Four uses of “observer”
In an ordinary conversation, an observer may be a person who sees an event. In a measurement description, it may mean the apparatus that responds to a signal. In a discussion of records, it may identify the system that retains a result. In relativity, it may refer to a worldline or reference frame used to assign coordinates and clock readings.
A reference frame does not need eyes. A detector does not automatically preserve everything that interacted with it. And a stored number becomes scientifically useful only when its meaning can be recovered: what was measured, under what conditions, with which calibration?
These distinctions help a discussion advance without prematurely deciding every question about consciousness. First identify the physical interaction and the available record. Then say what additional role, if any, the argument assigns to a conscious reader.
A one-bit observation
Consider a deliberately simple temperature instrument. At 10:00 it samples the room and stores one bit:
R(T) = 0 when T < 20°C; R(T) = 1 when T ≥ 20°C.
Suppose the stored result is 0. Under the specified calibration and sampling conditions, it tells us which side of the threshold the temperature occupied. It does not distinguish 10°C from 19.9°C. Those temperatures correspond to different physical conditions but the same displayed record.
This loss of detail need not be a defect. If the instrument's task is simply to trigger a threshold warning, the bit may be sufficient. If the task is to predict the time until the room reaches 20°C under a known heating rate, that same bit is insufficient. The adequacy of an observation depends on the question being asked.
A record has a history
At 10:05 the room has warmed above 20°C. The saved 10:00 bit remains 0. It is now a record of an earlier condition, not a claim that the present temperature is still below the threshold.
This simple example separates a physical state from the event it represents. The stored bit exists now; its reference concerns the earlier sampling. A usable record therefore needs more than a numerical value. It needs a reliable association with its source and, where relevant, a time label.
It also needs a way to distinguish “measurement returned 0” from “measurement did not occur.” If an uninitialized instrument displays 0 by default, the absence of a warning cannot establish a successful observation. A completion flag or documented acquisition procedure can carry information that the displayed result alone lacks.
Objectivity through comparison
Suppose two people read the stored bit and agree. Their agreement establishes that they read the same display, but it does not by itself validate the thermometer. Both may be consulting the same faulty record.
A stronger comparison examines calibration, sampling conditions and independent checks. Objectivity grows through procedures that allow different people to locate disagreement and reproduce the relevant measurement. It does not require an observer who stands outside the physical world.
In a model, the observer and its memory should therefore be included whenever they affect the proposed explanation. If a calculation says that information is retained, ask where it is retained and which later operations can use it.
Observation in the research programme
The SMONAD discussion of observers and records treats these questions as part of the relation between an underlying account and recoverable experience. A proposed description of reality must explain the stable distinctions available to an observer within it.
Our threshold instrument gives that broad requirement a modest test. Can the description represent sampling, retention and later reading as related events? Can it distinguish a valid result from a missing one? Can it explain why several temperatures produce the same record?
For quantum measurements, further issues arise about preparation, alternatives and outcome probabilities. The essay What does a quantum state describe? begins with those calculations. Keeping them explicit prevents “observer” from becoming a substitute for an explanation.
A question for any model
Choose one observation in a theory or experiment you are reading. Write down the input, the recording operation, the retained result and the later inference. Where does information disappear? Where is a time association supplied? Which assumptions connect the result to the event?
This exercise often reveals that the interesting question is not simply whether an observer exists. It is what that observer can reliably distinguish, remember and use.
Sources and further reading
- Observer and record: the related SMONAD philosophical guide.
- Memory for prediction: a worked example of information needed beyond a present display.
The threshold instrument is an original idealized example. Its exact cutoff and perfect retention are stipulated for clarity; a real instrument requires a measurement uncertainty and an account of its reliability.