Wait, what? A pressure sensor is tested at the same pressure twice. On the way up from a lower pressure it reads 50.3 units. On the way down from a higher pressure it reads 49.8 units. The laboratory did not change the target value. The display still changed. Is one reading automatically a mistake?
Not necessarily. Some measuring systems show hysteresis: their indication at a given input can depend slightly on the direction from which that input was approached. This is a powerful Reality Lab case because it forces us to notice something easy to ignore: scientific evidence can carry a history.
Quick Answer
A hysteresis specification tells us about a difference between readings obtained at the same input during increasing and decreasing runs under specified conditions. It does not mean that the sensor is simply “0.5% accurate”, and it does not mean every reading will be wrong by exactly 0.5%.
Your learner job is to ask: Was the same input reached by different paths, and were those paths compared fairly? If yes, the difference may be evidence about the measuring system rather than evidence that the physical input changed.
The Owned Learner Job
This article owns one narrow transfer job: evaluating a real-world sensor or calibration claim when the same input may give different indications depending on whether the input was increasing or decreasing.
It does not replace the existing eduKateSengkang guides on variables, fair testing, repeated readings, graph interpretation, measurement uncertainty, anomalous results or systematic shift. Those are the tools. Here, we use them on a communication object that says something like “Hysteresis: 0.5% FS.”
Rebuild the Evidence Object
Imagine an original sensor data sheet for a 0–100 unit instrument. It states: Hysteresis: ≤0.5% FS. “FS” means full scale. For this instrument, full scale is 100 units, so 0.5% of full scale is 0.5 unit.
| Target input | Approached from below | Approached from above | Difference |
|---|---|---|---|
| 20 | 20.1 | 19.9 | 0.2 |
| 50 | 50.3 | 49.8 | 0.5 |
| 80 | 80.2 | 79.9 | 0.3 |
The largest paired difference in this constructed example is 0.5 unit. That is the evidence object. Notice what it does not say. It does not say the true value is 50.05. It does not prove either direction is always closer to the reference. It does not describe every possible source of measurement error.
Observed, Claimed and Inferred
| Layer | What we may say |
|---|---|
| Observed | At the 50-unit reference, the increasing run indicated 50.3 and the decreasing run indicated 49.8. |
| Claimed | The product sheet states hysteresis is no more than 0.5% of full scale under its test conditions. |
| Inferred | The instrument may retain a path-dependent effect, so direction of approach matters when comparing measurements. |
A careful learner keeps these layers separate. A number printed on a specification sheet is a claim about performance under a defined method. Your job is to connect it to the method that produced it.
Why Direction Can Matter
Real measuring systems are physical systems. Parts can flex, rub, heat, cool, stick, relax or respond differently after different histories. NIST has documented hysteresis in real calibration work, including multi-hole Pitot sensors where increasing and decreasing conditions produced different responses in certain ranges. The important lesson for Primary Science is not the advanced engineering. It is the evidence habit: the route to a condition can sometimes matter to the reading at that condition.
This is why a calibration run that only increases the input may miss information that appears when the input is later decreased. A well-designed comparison can deliberately visit the same target from both directions.
Representation Check: Read the Arrows, Not Just the Dots
Suppose a graph has two curves: one for increasing input and one for decreasing input. A tempting response is, “There are two lines, so there must be two sensors.” That conclusion does not follow. The two lines may be two journeys made by the same sensor.
Look for arrows, run labels, sequence numbers or captions such as “upscale” and “downscale”. These small labels can carry the decisive meaning. Before comparing the vertical distance between curves, establish what each curve represents.
Comparison and Baseline Check
The phrase “0.5% FS” has a denominator hidden inside it: the instrument’s full-scale range. If the full scale is 100 units, 0.5% FS is 0.5 unit. If the full scale is 1,000 units, the same percentage corresponds to 5 units. So the percentage cannot be interpreted without the range.
Also check whether the manufacturer reports hysteresis as a maximum difference, a typical value or another convention. Do not silently convert one wording into another. In real science, definitions belong to the measurement method.
Method Check: What Would a Fair Hysteresis Test Need?
- The same reference points on the increasing and decreasing runs.
- The same instrument, units and environmental conditions.
- A controlled sequence so the direction of approach is known.
- Enough settling time that simple response delay is not mistaken for hysteresis.
- A reference standard suitable for the required comparison.
- Repeated cycles if we want to know whether the paired difference is reproducible.
These conditions matter because a difference between two readings has several possible explanations. Hysteresis is one. It is not the only one.
Alternative Explanations You Must Keep Alive
If the increasing and decreasing readings differ, consider at least these alternatives before deciding why: the sensor may not have settled; room temperature may have changed; the reference source may have drifted; the instrument may have low resolution; the operator may have read the scale differently; or the system may show genuine hysteresis.
Good scientific reasoning does not mean naming the fanciest explanation. It means asking which explanation is best supported by the pattern and the method.
Worked Case 1: A Spring Scale
A composite classroom-style case uses a spring scale. Masses are added until the reference force reaches a chosen value. Later, larger masses are removed until the same reference value is reached again. The pointer sits slightly higher on the first journey than on the second.
Weak reasoning: “The force must be different because the pointer is different.” Better reasoning: “The reference force is stated to be the same. The different indications therefore tell us something about the measuring system or the method. We should compare repeated up-and-down cycles before deciding the cause.”
Worked Case 2: A Humidity Sensor
A humidity chamber is held at 60% relative humidity after coming from 40%, and the sensor reads 60.4%. On another run the chamber is held at the same 60% after coming from 80%, and the sensor reads 59.9%.
The difference is 0.5 percentage point. It would be careless to call this “0.5% inaccurate” without checking the specification and reference. It would also be careless to average the two readings immediately and declare the average true. The important evidence is the paired difference at the same reference condition.
Worked Case 3: A Product Comparison
An advertisement compares Sensor A with “hysteresis ≤0.2% FS” and Sensor B with “hysteresis ≤0.8% FS”, then announces that Sensor A is “four times more accurate”. That conclusion travels too far.
Lower hysteresis can be an advantage for a job in which direction changes matter, but overall measurement performance can also depend on calibration, bias, repeatability, resolution, environmental effects, range and the exact use case. One performance characteristic does not automatically equal total accuracy.
Evidence That Would Strengthen the Hysteresis Interpretation
- The paired difference appears repeatedly at the same reference points.
- The sign or shape of the difference is linked consistently to direction of approach.
- Changing the order or repeating a full cycle preserves the pattern.
- Reference values and environmental conditions remain stable.
- Allowing adequate settling time does not remove the difference.
Evidence That Would Weaken It
- The difference disappears when enough settling time is allowed.
- The reference source was drifting during the run.
- Only one paired point was measured once.
- The instrument was moved, warmed or reconfigured between runs.
- The apparent gap is smaller than the display resolution.
How Far Can the Conclusion Travel?
A hysteresis test on one instrument, one range and one set of environmental conditions supports conclusions about that tested situation. It does not automatically prove the same behaviour for every unit of the same model, every temperature, every rate of change or every point outside the tested range.
This is a core PSLE Science habit: match the width of the claim to the width of the evidence.
Tempting but Invalid Reasoning
| Tempting statement | Why it fails |
|---|---|
| “Same input means identical reading every time.” | Real measuring systems can show path-dependent behaviour and other variation. |
| “0.5% hysteresis means 99.5% accurate.” | Hysteresis is one performance characteristic, not total accuracy. |
| “Average the two directions and the problem disappears.” | Averaging can hide the directional pattern instead of explaining it. |
| “The lower reading must be the correct one.” | Direction alone does not identify truth; a reference is needed. |
PSLE-Style Transfer Case
An original investigation uses a force sensor. During Run P, the force is increased from 0 N to 10 N in 2 N steps. During Run Q, it is decreased from 10 N to 0 N in 2 N steps. At the 6 N reference point, the sensor reads 6.2 N in Run P and 5.9 N in Run Q.
A strong response is: The two readings were taken at the same reference force but after the force had changed in opposite directions. The difference therefore may be related to the sensor’s response to the direction of approach. More repeated up-and-down cycles under the same conditions would be needed to determine whether this pattern is consistent.
Notice what the answer does not do. It does not pretend to know the internal mechanism. It does not invent a mark-scheme keyword. It uses the evidence supplied and keeps uncertainty visible.
Delayed Independent Return
Tomorrow, without rereading this page, answer three questions: (1) Why can the same reference input give two different indications in a hysteresis test? (2) Why is a hysteresis percentage not the same as overall accuracy? (3) What observation would help separate hysteresis from simple response delay?
Explained Practice
- A gauge reads 30.2 when approached from 20 and 29.9 when approached from 40. What is the evidence-bearing comparison? Answer: the paired readings at the same 30-unit reference, with opposite directions of approach.
- A product has lower hysteresis than another. Can we call it more accurate in every way? No. We need other performance evidence and the intended use.
- Increasing and decreasing runs differ only when the operator changes quickly. What alternative should be checked? Response or settling time before concluding that the difference is persistent hysteresis.
Route Back to the Canonical PSLE Science Skills
Reality Lab applies existing skills rather than taking ownership of them. For deeper work, return to reading repeated results, random variation versus systematic shift, and controlled conditions that quietly change.
Parent and Tutor Teaching Guide
Do not begin by asking a child to memorise the word hysteresis. Begin with two journeys to the same destination. Draw 20 → 50 and 80 → 50. Give two readings at 50. Ask, “What stayed the same? What was different before the reading was taken?” Once the learner sees that the history differs, attach the technical term.
Then ask the learner to design one check that would make the evidence stronger. Good answers include repeating the cycle, controlling temperature or waiting long enough for the sensor to settle. The teaching goal is not jargon. It is evidence discipline.
Authoritative Sources
- NIST: The Effect of Turbulence on a Multi-Hole Pitot Calibration — a real calibration example in which hysteresis was observed over particular conditions.
- NIST: Hysteresis Compensation in Temperature Response of Fiber Bragg Grating Thermometers — demonstrates that hysteresis can contribute to measurement uncertainty.
- Singapore Examinations and Assessment Board: PSLE and the current 2026 PSLE Science syllabus — official examination context for applying knowledge and scientific inquiry.
The Quiet Habit
When two readings disagree at the same stated input, do not rush to choose a winner. First reconstruct the route that produced each reading. Sometimes the missing evidence is not another decimal place. It is the direction from which the measurement arrived.
