Wait, what? A laboratory sends a thermometer for calibration. The report comes back with two lines: As Found: Out of Tolerance and As Left: In Tolerance After Adjustment. Someone smiles and says, “Good. The thermometer is correct now, so all the measurements we took last month are fine.”
That last step does not follow automatically. A successful adjustment today changes the instrument’s present condition. It does not travel backward in time and rewrite what the instrument was doing before it was adjusted.
Quick Answer
As found describes the condition measured before repair or adjustment during the calibration visit. As left describes the condition after any adjustment or other permitted work. When judging measurements made before the calibration, the as-found information is usually the more directly relevant evidence about the instrument’s state at the end of that earlier use period.
Even then, do not jump from “out of tolerance as found” to “every earlier result was wrong.” The evidence may show a risk to earlier measurements, but deciding how much earlier data are affected requires more information: which points were out, by how much, how the instrument was used, whether check standards or comparisons existed, and whether drift was gradual or sudden.
The Owned Learner Job
This Reality Lab owns one narrow transfer job: reading a calibration report as a time-linked evidence object. The learner must connect “as found” and “as left” to the correct part of the measurement story instead of treating a calibration certificate as one timeless stamp of correctness.
It does not replace existing PSLE Science owners for measurement, checking, repeated readings, uncertainty, conclusions or alternative explanations. It applies those tools to a real communication object used by laboratories, technicians and measurement systems.
Rebuild the Calibration Story
Consider an original composite case. A digital thermometer was used in a school science room from January to June. On 1 July it was sent for calibration against suitable reference temperatures.
| Reference point | As-found indication | Difference | After adjustment |
|---|---|---|---|
| 0.0°C | 0.1°C | +0.1°C | 0.0°C |
| 50.0°C | 50.8°C | +0.8°C | 50.1°C |
| 90.0°C | 91.2°C | +1.2°C | 90.1°C |
Suppose the allowed tolerance is ±0.5°C. The as-found results are within tolerance at 0°C but outside tolerance at 50°C and 90°C. After adjustment, the instrument is within tolerance at all three tested points.
What does this prove? It supports the statement that the thermometer had an unacceptable positive error at some tested points when it arrived for calibration, and that after adjustment its tested performance improved. It does not by itself tell us exactly when the earlier error began or what the error was at every temperature on every previous day.
Three Time Windows, Three Different Questions
| Time window | Evidence question |
|---|---|
| Before calibration | What condition was the instrument likely in while earlier measurements were being made? |
| During calibration | What did comparison with references reveal before and after adjustment? |
| After calibration | What can we say about the instrument’s tested condition now, under the stated calibration conditions? |
A learner who mixes these windows can make a serious reasoning error. “It passed after adjustment” answers a question about the instrument after work was done. It does not automatically answer the historical question.
Observed, Claimed and Inferred
Observed: the calibration comparison produced specified as-found and as-left differences at tested reference points.
Claimed: the report says the instrument was out of tolerance as found and in tolerance as left.
Inferred: some measurements taken before calibration may need review because the instrument was not meeting its tolerance when tested before adjustment.
The inference is reasonable, but its size is not yet known. That is where scientific judgement begins.
Why “Out of Tolerance” Does Not Mean “Every Old Number Is False”
A tolerance is a decision boundary. Crossing it tells us the instrument did not meet a stated requirement at a tested point. But earlier measurements may have been made at different parts of the range, with different required tolerances, or alongside independent checks. If the as-found error was large only near 90°C, a result measured near 10°C may need a different review from one measured near 85°C.
The right response is not panic and not dismissal. It is a targeted evidence review.
Method Check: Questions That Matter Before Judging Old Data
- Which calibration points were outside tolerance?
- How large and in what direction were the differences?
- What range of values did the earlier experiments actually use?
- Were there intermediate checks, reference materials or comparison instruments?
- Was the instrument dropped, repaired, overheated or otherwise changed during the period?
- Is there evidence of gradual drift or only one end-of-period comparison?
- How large an error would be important for the scientific conclusion being made?
Representation Check: A Certificate Is Not a Time Machine
A certificate often looks authoritative: dates, serial numbers, tables, signatures, standards and uncertainty statements. That authority is useful, but the document still has a scope. NIST notes that calibration or test results apply to the specific instrument or standard at the time of test unless otherwise stated.
This one sentence is a major scientific habit. Every evidence object has a time, object and condition attached to it. Remove those, and a precise statement can turn into an overclaim.
Worked Case 1: The Seed-Germination Incubator
A class used a thermometer to keep an incubator near 30°C. Two months later the thermometer was found to read 0.7°C high at a 30°C reference. After adjustment it reads correctly.
Invalid conclusion: “The old experiment is useless.” Better conclusion: “The as-found check suggests the earlier displayed temperatures near 30°C may have been higher than the reference by about the observed amount at the time of calibration. We should review whether a difference of this size could change the experiment’s interpretation and look for independent temperature records.”
Worked Case 2: The Balance
A balance was used to compare two soil samples with masses around 50 g. At calibration it is as found within tolerance at 50 g but outside tolerance at 500 g. An alarmed student says every 50 g result must be discarded.
The available evidence does not support that blanket action. The out-of-tolerance point is far from the range used in the experiment. That does not prove the 50 g measurements were perfect, but the relevant as-found result at 50 g matters more directly. Evidence should be matched to the region of use.
Worked Case 3: “Freshly Calibrated” Advertising
A product comparison says, “Device A was freshly calibrated, so its data from last year are more trustworthy.” That claim confuses time. A current calibration can support present measurement confidence when used appropriately. It does not automatically certify measurements made before the current calibration.
Evidence That Strengthens Concern About Earlier Measurements
- The as-found error is large in the same range used for the earlier work.
- Independent check standards also show growing disagreement before calibration.
- The instrument suffered an event that could plausibly shift its response.
- The scientific conclusion depends on differences smaller than the observed as-found error.
- Several tested points show a consistent directional bias.
Evidence That Limits or Weakens That Concern
- The earlier work used a range where as-found performance remained within requirement.
- Independent references recorded agreement during the earlier period.
- The scientific conclusion is much larger than the possible measurement shift and remains unchanged after a reasonable correction check.
- The out-of-tolerance condition is linked to a later event that occurred after the earlier data were collected.
Tempting but Invalid Reasoning
| Tempting statement | Why it fails |
|---|---|
| “As left is good, so the past was good.” | As-left evidence describes the instrument after adjustment. |
| “As found is bad, so every old result is false.” | The effect on earlier results depends on range, magnitude, timing and independent checks. |
| “Calibration fixes data.” | Calibration and adjustment act on the instrument; historical data require separate review. |
| “One failed point tells us the error everywhere.” | Performance can vary across the range. |
How Far Can the Conclusion Travel?
A calibration report supports statements about the specific item, reference points, method, date and stated conditions. Do not turn that into a universal statement about every past or future measurement. NIST explicitly cautions that calibration results apply to the specific instrument at the time of test unless otherwise stated.
PSLE-Style Transfer Case
An original data logger measured water temperature near 40°C during an investigation in May. In June, before adjustment, a calibration check found that the logger read 40.6°C when the reference was 40.0°C. After adjustment it read 40.0°C.
A strong evaluation is: The as-found result is relevant to the measurements made before adjustment because it shows the logger was reading high at the tested 40°C point when checked. However, the June result alone does not prove the logger had the same 0.6°C error throughout May. Earlier check data or another reference would strengthen the conclusion about the May measurements.
This answer uses evidence and uncertainty together. It neither ignores the calibration finding nor turns it into more certainty than it deserves.
Delayed Independent Return
Tomorrow, draw a three-box timeline labelled before calibration → calibration → after adjustment. Place “as found” and “as left” in the correct positions. Then explain which one is more directly relevant to data collected before the calibration visit and why.
Explained Practice
- A meter is out of tolerance as found at 100 units but within tolerance at 10 units. Earlier work used only 8–12 units. What should you do? Focus first on the evidence near the range actually used; do not automatically transfer the 100-unit failure to 10 units.
- An instrument is adjusted and passes as left. Does that validate last month’s readings? No. It supports its condition after adjustment, not historical validation by itself.
- What would strengthen a claim about earlier data? Intermediate check standards, comparison instruments, maintenance records or evidence showing when the shift occurred.
Routes to Existing PSLE Science Owners
Use this page to transfer existing skills, then revisit random variation versus systematic shift, repeated measurements, and checking whether a conclusion goes beyond the evidence collected.
Parent and Tutor Teaching Guide
Use a simple clock story. Imagine a clock that is discovered at noon to be five minutes fast, then corrected. Ask: “Does correcting it at noon prove every time written down yesterday was exact?” Most learners immediately see the time logic. Then transfer the same structure to a thermometer or balance.
Next, add the harder question: “Does finding it five minutes fast at noon prove it was exactly five minutes fast all day yesterday?” The answer is also no. Now the learner can see both boundaries: current correction does not repair the past, and one current error measurement does not reveal the entire past history.
Authoritative Sources
- NIST Calibration Policies — explains that calibration/test results apply to the specific instrument or standard at the time of test unless otherwise clearly stated.
- NIST Metrological Traceability FAQ and Policy — explains the scope and limits of traceability claims and later user measurements.
- SEAB 2026 PSLE Science syllabus — current objectives include evaluating observations, information and methods and communicating explanations and reasoning.
The Quiet Habit
Whenever a report contains “before” and “after” evidence, pin each result to its place on the timeline. A repaired instrument can become better evidence for tomorrow. It cannot, by itself, travel backward and certify yesterday.
