PSLE Science Reality Lab Vol No.095 | “The Calibration Curve Stops at 100” — Can We Trust a Reported Value of 140?

PSLE-SCI-REALITY-0095

Wait, What? A straight line drawn on a graph can keep going long after the evidence has stopped.

A fictional sensor is calibrated using known standards of 0, 20, 40, 60, 80 and 100 units. The points form a neat rising line.

Then an unknown sample gives a much larger signal. Software extends the line beyond the last standard and reports: 140 units.

The number looks scientific. The graph looks tidy. But there is an important evidence question: was the measuring relationship actually tested at 140?

Reality Lab Vol No.095 teaches a durable transfer habit: when a measurement is inferred outside the range covered by reference standards, distinguish a mathematical extension from a tested measurement relationship.

Quick Answer

Find the lowest and highest reference standards used to establish the calibration.
Locate the unknown relative to that supported range.
Distinguish interpolation inside the tested range from extrapolation beyond it.
Ask whether the instrument response is known to keep the same relationship outside the tested region.
Do not assume a straight line must remain straight forever.
Prefer an in-range measurement, a wider validated calibration, or an appropriate sample adjustment where the method allows it.
Report the value only as strongly as the calibration evidence supports.

What This Page Owns — and What It Leaves With Existing Owners

This is not a full calibration lesson or a specialist analytical-chemistry guide. It applies existing measurement reasoning to one real-world communication object: a reported scientific value that comes from extending a calibration relationship beyond the region actually covered by standards.

Original Reality Lab Case: The Purple-Signal Meter

This is an original teaching case using fictional measurements. It does not reproduce a real commercial assay or examination question.

A meter measures the intensity of a fictional purple signal. Known standards give the following results:

Known amount	Meter signal
0	0.2
20	1.9
40	3.8
60	5.7
80	7.6
100	9.5

The standards look almost perfectly linear from 0 to 100. The unknown gives a signal of 13.3. Extending the same line would suggest roughly 140 units.

But the experiment has not shown what the meter does between 100 and 140. Perhaps it remains linear. Perhaps the detector begins to saturate. Perhaps the chemistry changes. Perhaps the signal curve bends. The extrapolated number is a prediction based on a model, not a direct demonstration that the model remains valid there.

Observed, Modelled and Claimed

Layer	Statement
Observed standards	Known values from 0 to 100 produced a near-linear response.
Observed unknown	The unknown produced a signal larger than the 100-unit standard.
Modelled	Extending the fitted line maps that larger signal to about 140 units.
Claimed	The unknown has been reliably measured as 140 units.

The claim is stronger than the direct calibration evidence unless the method has additional evidence showing that this extension is valid.

Interpolation and Extrapolation

Interpolation estimates a value inside the region covered by known standards. Extrapolation extends the relationship beyond that region.

For Primary 5/6 reasoning, the key idea is not the vocabulary itself. It is the evidence boundary. Inside the calibration range, the unknown is surrounded by tested reference points. Outside it, the learner is assuming the relationship continues into territory that was not covered by those standards.

Why a Straight Line Might Stop Being Straight

A detector can approach its maximum response.
A sensor can become less sensitive at high values.
A chemical or physical response can change shape.
The instrument can clip or saturate.
Background effects can become more important.
The model may have been chosen only because it fits the tested region.

None of these possibilities proves the 140-unit estimate is wrong. They explain why it needs more evidence.

The Range Is Part of the Claim

A calibration is not just a line. It is a relationship established using reference standards, an instrument, conditions, a method and a range of interest. NIST calibration guidance describes standards selected to cover the instrument or application range and then uses that calibration relationship for future measurements.

That makes the endpoints scientifically meaningful. They tell you where the relationship was actually anchored by known values.

Worked Case 1: Unknown at 70

The standards cover 0–100 and the unknown corresponds to about 70. The estimate lies inside the reference range. That does not make the measurement perfect, but the calibration directly covers the region around the unknown.

Worked Case 2: Unknown at 140

The standards still stop at 100, but the unknown maps to 140 only by extending the line. The learner should label that step as extrapolation and ask for evidence that the same response relationship remains valid above 100.

Worked Case 3: Bring the Unknown Back Into Range

In some measurement methods, a sample can be adjusted in a documented way so that the measured response falls inside the calibrated region, and the original amount can then be calculated using that known adjustment. Whether that is appropriate depends on the method. The important reasoning move is that the actual measurement occurs inside the supported range rather than relying blindly on the extended line.

Worked Case 4: Extend the Calibration With New Standards

Another approach is to include suitable reference standards above 100 and test whether the relationship really remains valid. If 120, 140 and 160 standards confirm the response model and measurement quality, the evidence boundary moves outward.

Worked Case 5: The Curve Bends

Suppose new standards at 120 and 140 produce signals lower than the straight-line prediction. The earlier extrapolation overestimated the amount because the response began to level off. Nothing mysterious happened: the graph extension was a model assumption that the new standards tested.

The Graph Can Continue Even When the Evidence Does Not

Graphing software makes it easy to extend trend lines across an entire page. That visual continuity can hide a scientific discontinuity: the model may be supported by reference data only over one section.

A careful graph therefore makes the calibration region visible or explains when values are extrapolated.

What Evidence Would Strengthen the 140 Claim?

Reference standards that extend through and beyond 140.
Evidence that the instrument remains responsive and unsaturated in that region.
A validated model that is appropriate over the wider range.
Repeated measurements with acceptable variation around 140.
An independent or alternative in-range measurement that agrees.
A documented sample adjustment that returns the measurement to the validated range.

What Would Weaken It?

The highest reference standard is 100 and no higher standards were tested.
The unknown signal is beyond the instrument’s stated range.
The detector is known to saturate near the top end.
A linear line was chosen for convenience rather than validated behaviour.
No uncertainty or warning is given for the extrapolated region.
A precise-looking number is reported without saying it lies outside calibration.

Tempting Reasoning That Fails

“The line is straight from 0 to 100, so it must stay straight forever.” The tested data establish the relationship only over their supported region unless wider evidence exists.
“The software produced 140, so 140 was measured.” Software can extrapolate a model beyond measured standards.
“Any extrapolation is automatically wrong.” No. It is a prediction beyond the observed range and can be defensible when supported by theory or validation. It simply carries a different evidential status.
“The last standard is a magical physical limit.” Not necessarily. It marks the edge of this calibration evidence, which can sometimes be expanded with appropriate standards.

How Far Can the Conclusion Travel?

If the unknown is outside the calibrated region, a cautious report might say: “The signal exceeds the highest calibration standard; a reliable amount requires a method that brings the result within a supported range or extends the calibration.”

That sentence is scientifically useful. Refusing to invent precision is not failure; it is evidence control.

PSLE-Style Transfer Case

A fictional sensor is calibrated with known values 0, 25, 50, 75 and 100. An unknown gives a response larger than the 100 standard. A straight-line extension suggests 135.

Question: Why should the learner hesitate before writing “unknown = 135 units”?

Reasoned answer: The calibration standards only tested the response relationship up to 100. The 135 value comes from extending the model beyond the tested calibration range. A wider calibration or another method that measures the unknown within a supported range would strengthen the numerical claim.

Explained Practice

Practice A: Standards cover 10–50 and an unknown estimates to 35. Interpolation or extrapolation? Interpolation, because 35 lies within the covered range.

Practice B: Standards cover 10–50 and the unknown estimates to 72. What should you ask? Whether the calibration relationship is validated above 50, and whether the sample can be measured using a supported in-range approach.

Practice C: New standards at 60, 70 and 80 confirm the same relationship. What changed? The evidence range has expanded.

Delayed Independent Return: The R-A-N-G-E Check

R — References: Which known standards built the calibration?
A — Area: Is the unknown inside or outside that range?
N — Nature of the model: Why should the relationship continue?
G — Guard against saturation: Can the instrument still respond properly?
E — Earn the number: Use in-range or extended validation evidence before claiming it strongly.

Parent and Tutor Teaching Guide

Draw five known points from 0 to 100 and a straight fitted line. Ask the learner to estimate a value at 60, then at 140. Both can be read mathematically, but only one lies among the tested standards.

Then add new points at 120 and 140 that bend away from the old line. This makes the core habit visible: a model can travel farther on paper than the evidence that built it.

Authoritative Sources

NIST calibration guidance describes reference standards selected to cover a measurement range and a model established from those standards. Older NIST measurement guidance also warns that extrapolation outside the range used to compute a fitted relationship can lead to serious error. The Primary Science lesson is simple: mark where the evidence ends before extending the line.

The Quiet Return

A line can be longer than the experiment.

Before trusting the number beyond the last standard, ask what evidence lives there.

Small Group Tutorials