How to Tell Whether a Measurement Error Shifts Both PSLE Science Set-Ups Together or Creates a False Difference

Wait, What? A measuring error can make every reading wrong without making the difference between two readings wrong by the same amount.

Suppose a measuring instrument reads 2 units too high every time. A true value of 10 appears as 12. A true value of 15 appears as 17. Both absolute readings are wrong, but the measured difference is still 5 units. Now imagine the error affects only one set-up: 10 appears as 12 while 15 remains 15. The apparent difference becomes 3 units. The scientific comparison has changed.

This is a powerful PSLE Science reasoning distinction. “There is a measurement error” is not yet enough. You must ask which readings are affected, in what direction, and whether the comparison itself changes.

Quick Answer

Separate two questions. First: Are the absolute readings accurate? Second: Does the error change the comparison between the set-ups? A shared offset can shift both readings together and leave a simple difference unchanged. An unequal, one-sided or condition-dependent error can create, hide, reverse or exaggerate a difference. Even when a difference survives, do not claim the absolute values are accurate.

The PSLE Science Learning Job This Guide Owns

This guide owns one learner job: judging how a measurement error propagates into a comparison. It does not replace the broader owner on systematic and random error, instrument range, resolution or calibration. It applies those ideas specifically to PSLE Science comparisons: shared error versus differential error.

SEAB’s PSLE Science assessment objectives from 2026 include interpreting and analysing information and evaluating observations, information and methods. A learner therefore needs to know not only that a method is imperfect, but what that imperfection does to the evidence.

The Two-Layer Check

Layer	Question
Absolute-value layer	Could the recorded values all be shifted or biased away from the true values?
Comparison layer	Does that bias affect both set-ups equally, or does it change one more than the other?

Do not collapse these layers. A comparison can sometimes remain informative even when the absolute readings are biased. Conversely, readings can look precise while a one-sided bias creates a misleading comparison.

Worked Example 1 — Same Offset on Both Set-Ups

A fictional instrument reads 3 units too high throughout an investigation. Set-up P has a true value of 20 and Set-up Q a true value of 26. The instrument shows 23 and 29.

The displayed values are not accurate.
The displayed difference is 6 units.
The true difference is also 6 units.

The common offset moves both readings together. It does not repair the inaccurate absolute values, but in this simplified example it preserves the difference.

The lesson is not “systematic error does not matter”. It matters whenever the actual value matters, when the offset is unknown, when thresholds matter, when the error is not truly equal across the range, or when another calculation depends on the absolute reading. The narrower lesson is that the effect of an error depends on the claim you are making.

Worked Example 2 — Error in One Set-Up Only

P should read 20 and Q should read 26. A method problem adds 3 units only to P, so the recorded values are 23 and 26.

The apparent difference is now 3 instead of 6. The error has not merely shifted the scale; it has changed the comparison. If the error were large enough, it could even make the lower true value appear higher.

Worked Example 3 — Same Instrument, Different Effect

Using the same instrument for both set-ups does not guarantee the same measurement error. One sample may be outside the instrument’s useful range, one may be harder to align, one may be measured at a different angle, or the method may interact differently with the two conditions.

Therefore “same instrument” is a useful fairness feature, but it is not proof that every error is common and equal. Read the method details.

Worked Example 4 — Shared Rounding

Suppose two values are recorded to the nearest whole unit. True values of 10.4 and 10.6 may appear as 10 and 11. Here the effect is not a simple fixed offset. Rounding can change each value differently depending on where it lies relative to a boundary.

This is why you should not assume every common measuring method produces a perfectly common error. The error structure matters.

Worked Example 5 — A Shared Error Can Hide a Threshold Problem

If a scientific decision depends on whether a reading is above or below a threshold, a common +2-unit shift can matter even if the difference between P and Q is preserved. Both values may cross the threshold in the display even when one or both true values do not.

Again, always ask what claim the evidence must support: exact value, ranking, difference, threshold, trend or something else.

Common Shift, Unequal Shift or Random Scatter?

Pattern	What it can do	Useful learner question
Same fixed shift on all readings	Absolute values wrong; some differences may be preserved	Does my claim depend on the actual values or only their difference?
Unequal shift between set-ups	Can create, hide or change a difference	Why would one reading be affected more?
Random scatter across repeats	Can make the observed difference unstable	Do repeats show the difference consistently?
Instrument limit or saturation	Can make different true values look the same	Has the method stopped distinguishing the set-ups?

Failure Signatures

“There is an error, so the comparison is useless.”
“The same instrument was used, so both errors must be identical.”
“Both readings are shifted, therefore the absolute values are still correct.”
“Repeating the measurement will remove a fixed bias.”
The learner notices an instrument problem but never states what conclusion it affects.
A one-sided method problem is treated as a harmless shared offset.

Earliest Weak-Link Diagnosis

State the claim: exact value, difference, ranking, threshold or trend?
Identify which readings may be affected.
State the direction of the possible error if known.
Ask whether the effect is equal across the compared set-ups.
Rebuild the comparison under that error pattern.
Limit the conclusion if the error cannot be quantified.

Misconception Repair

“Systematic means the same error in every circumstance.” Not necessarily. A systematic bias is a consistent tendency, but its size can depend on the method, range and condition. Do not invent a constant offset unless the question supports it.

“If the difference survives, the experiment is accurate.” No. Preserving one comparison does not make the absolute readings accurate.

“Repeats fix every error.” Repetition can reveal variation, but repeating a biased method can reproduce the same bias.

“Any error means no conclusion is possible.” Sometimes the method still supports a limited comparison. Science asks what remains justified, not whether the method is perfect.

The Comparison-Error Protocol

NAME THE CLAIM → IDENTIFY THE POSSIBLE ERROR → ATTACH IT TO THE AFFECTED READINGS → ASK SHARED OR UNEQUAL? → REBUILD THE COMPARISON → CHECK THRESHOLDS/RANGE → STATE WHAT SURVIVES → STATE WHAT DOES NOT.

Original Practice Set

Case A: P and Q display 12 and 20. The instrument is known to read exactly 2 units too high for both. What remains true? Their true values are 10 and 18; the 8-unit difference survives, while the displayed absolute values are high.

Case B: P and Q display 12 and 20, but only P may have an extra +2-unit error. Can the apparent 8-unit difference be trusted exactly? No. The true difference could be different.

Case C: Both readings hit the top of a scale. Is that a shared offset? No. It may be saturation; the instrument may no longer distinguish the true values.

Case D: Three repeats scatter around each set-up’s average. Is that automatically a fixed bias? No. The pattern may instead reflect random variation or another source of variability.

Retrieval and Transfer Sequence

Sort error descriptions into shared shift, unequal shift, random scatter or method limit.
For each, state whether absolute values, differences or both are threatened.
Change the scientific context but keep the same numerical structure.
Remove the numbers and reason qualitatively about which comparison can survive.
Return after several days and diagnose a new method without prompts.

Delayed Independent Return Test

Give the learner a fresh two-set-up investigation with a stated measurement problem. The learner passes if they do not stop at “there is an error”, but instead identify which readings are affected, whether the comparison changes and what conclusion remains defensible.

Answer-Checking Receipt

I know what claim I am evaluating.
I know which readings the error affects.
I have not assumed equal error without evidence.
I separated accuracy of absolute values from accuracy of the comparison.
I checked whether the error could create, hide or reverse a difference.
I checked range, threshold and rounding effects when relevant.
I did not claim repeats remove a fixed bias.
My conclusion says what survives and what does not.

Parent and Tutor Teaching Guide

Use two number cards and physically add the same amount to both. Ask what changed and what stayed the same. Then add an amount to only one card. The child can see immediately that “both wrong” and “comparison wrong” are different questions.

Next remove the exact error size. Say only, “P may be affected more than Q.” Ask the learner to stop making exact claims. This teaches the evidence boundary: sometimes you can diagnose a threat without being able to calculate the corrected value.

Keep the language age-appropriate. Terms such as common-mode and differential bias are useful in advanced science, but Primary learners only need the reasoning: same shift or different shift?

Useful Internal Routes

Authoritative References

Quiet Return

A method problem matters because it changes what the evidence can support. Do not stop at the word “error”. Follow the error into the readings, then into the comparison, then into the conclusion. That is how a Primary learner begins to think like a careful experimental scientist.

Small Group Tutorials