Measurement relates an indication to a quantity through a procedure, a model and a calibration. It makes claims about reality more precise when those connections and their uncertainties are explicit. A displayed number is a starting point for that account.
A reading is not yet the whole measurement
A sensor may display a voltage, a count or a position on a scale. To interpret that indication as temperature, duration or distance, we need a relationship between the readout and the quantity being investigated. The relationship can depend on the instrument, its environment and the way it is used.
The International Vocabulary of Metrology distinguishes a measurement model from the quantities entering it. It also treats calibration as establishing how indications are related to reference quantities with their uncertainties. This language helps make an inference inspectable rather than treating the instrument as an unexplained source of numbers.
An explicit two-point calibration
Suppose an idealized sensor follows the linear model y = ax + b. Here y is the raw indication and x is the quantity we want, expressed in specified units. A reference with x = 0 gives y = 1; a reference with x = 10 gives y = 21.
Under the linear assumption, the offset is b = 1 and the gain is a = 2. A later reading y = 9 therefore corresponds to:
x = (y − b)/a = (9 − 1)/2 = 4.
Reporting the raw 9 as the measured value of x would confuse the indication with the inferred quantity. Reporting 4 without mentioning the calibration could hide the assumption that makes it meaningful. The numbers are deliberately simple so that the role of each step remains visible.
Uncertainty has a source and a model
If the indication has an uncertainty of ±0.5 and the gain and offset are treated as exact, the corresponding interval for x is ±0.25. This follows from dividing the readout interval by the gain. It does not include uncertainty in either reference or any instability of the sensor.
Real calibration parameters have uncertainties too, often with correlations. Their influence must be propagated through the measurement model. Printing extra decimal places does not remove these contributions. The simplified interval calculation demonstrates one contribution, not a complete uncertainty budget for a physical instrument.
It is also necessary to state what the uncertainty expression means. An interval bound and a standard uncertainty derived from a probability model are not automatically the same thing. The example above uses a stated readout interval.
A check beyond the calibration points
At a third reference value, x = 5, the fitted model predicts y = 11. Suppose a new reading gives 12. That discrepancy prompts a check: perhaps the relation is nonlinear, the settings have changed, the reference is uncertain or the readout fluctuates.
The discrepancy alone does not identify the cause. Its significance depends on the uncertainties and measurement conditions. Adjusting the gain separately for every new reading would conceal the question of whether one calibration explains all the observations.
This connects measurement with the distinction between fitting and explanation. Calibration supplies a usable relation; additional observations examine whether that relation remains adequate in the setting where it is applied.
Interaction does not make the result arbitrary
A measurement is a physical interaction or comparison. It can alter what is measured, and a good model includes any alteration relevant to the result. That fact does not make the result a matter of personal choice. Reproducible procedures and comparisons between instruments constrain the interpretation.
Nor does the involvement of an observer establish that consciousness creates the measured property. Such a claim would require its own argument and evidence. The immediate task is to identify preparation, interaction, readout and the law relating them.
The physical dictionary in SMM/DTT
The expanded book's reconstruction requirements include clocks, units and observables. A count of inner operations is not already a duration in seconds, and a mathematical response is not already an instrument reading. A physical dictionary must connect them through shared, testable calibrations.
This is a constructive opportunity for the programme. If one specified model predicts several calibrated comparisons with common inputs, it explains more than a collection of separately adjusted numbers. Measurement is where the proposed connection between inner generation and completed observations becomes accountable.
Sources and further reading
- JCGM, International Vocabulary of Metrology, 2.39: Calibration and 2.48: Measurement model.
- Mohamed Haj Yousef, What a reconstruction must contain and A model that can be audited.
- Continue with Observation and inference. The numerical calibration is an original teaching example.