PSLE Science Reality Lab Vol No.007 | “Look How Much It Improved” — What Happens If the Time Window Changes?

Series ID: PSLE-SCI-REALITY-0007

Wait, What? The Same Data Can Look Like a Success, a Failure or Almost No Change

Imagine a graph showing the height of seedlings over twelve days. The headline says, “Special Treatment Makes Plants Grow Rapidly.” The graph shown in the post starts on Day 6 and ends on Day 9. During those four days, the treated seedlings rise sharply.

Now imagine that the complete record begins on Day 1 and ends on Day 12. From Day 1 to Day 5, the treated plants grew no faster than the comparison plants. From Day 6 to Day 9, they did. From Day 10 to Day 12, both groups levelled off.

Nothing in the four-day graph has to be fake. The measurements can all be genuine. Yet the claim can still become much stronger or weaker depending on which part of time is shown.

Primary Science trains exactly the habits needed here. The current syllabus asks learners to interpret and analyse information, evaluate observations and methods, communicate reasoning, use evidence carefully and exercise healthy scepticism about assumptions and uncertainty. The 2026 PSLE Science assessment objectives likewise include evaluating information and methods, not merely extracting a visible trend.

Quick Answer

When a claim depends on change over time, never inspect only the impressive rise or fall. Ask:

Why did the record begin here?
Why did it end here?
What happened before the displayed window?
What happened after it?
Did the measurement method or conditions change during the period?
Would a different reasonable start and end point tell the same story?

A selected time window is not automatically misleading. Sometimes a short window is scientifically appropriate. The learner’s job is to decide whether the window matches the question or quietly creates the conclusion.

The Owned Learner Job

This Reality Lab owns one transfer job: evaluating a real-world claim when the apparent improvement, decline or trend depends on where the time window begins and ends.

It does not replace the existing eduKateSengkang guides on graph reading, data gaps, trend interpretation, temporary versus sustained change, prediction beyond a tested range or how far a conclusion can travel. Those pages own the individual skills. This page puts them together around a common real-world communication pattern: the carefully selected slice of time.

Reality Lab Case: The Amazing Filter

Suppose an original advertisement for a classroom air-filter experiment says:

“Particles in the room fell by 60% after the filter was switched on.”

The accompanying chart shows particle readings at 2:00 pm and 2:30 pm. The filter was switched on at 2:00 pm. The number fell from 100 units to 40 units.

That sounds powerful. But a scientist should immediately want the wider record.

Imagine the full measurements were:

1:00 pm — 155
1:30 pm — 130
2:00 pm — 100
2:30 pm — 40
3:00 pm — 38
3:30 pm — 42

The room was already improving before 2:00 pm. That does not prove the filter had no effect. It means the two-point comparison alone cannot tell us how much of the change should be attributed to the filter.

Three Different Questions Hidden Inside One Graph

Question 1: Did the measured value change?

Yes. From 2:00 pm to 2:30 pm, the recorded value fell from 100 to 40. That is an observation about the selected interval.

Question 2: Was the selected interval unusual compared with the earlier trend?

Possibly. The earlier readings were also falling. We would need to compare the rate and pattern, not simply notice that both numbers decreased.

Question 3: Did switching on the filter cause the whole decrease?

The time series alone does not establish that. Other conditions may have changed. A comparison room, a repeated on/off design, or other controlled evidence could help distinguish the filter effect from the pre-existing decline.

The important habit is to keep change, trend and cause as separate jobs.

A Time Window Has Boundaries, and Boundaries Are Scientific Choices

Every time series begins somewhere and ends somewhere. Those choices are sometimes obvious: the experiment starts when the apparatus is activated and stops after a fixed duration. In other cases the boundaries are less natural. A post may begin at the highest value and end at the lowest. A product comparison may start just before an unusual improvement. A news graphic may stop before the trend reverses.

A learner should therefore ask whether the chosen window is:

mechanism-based — tied to a real event or process stage;
method-based — chosen before data collection, such as a fixed 30-minute interval;
question-based — appropriate to the exact scientific question;
or result-based — selected because it makes the effect look especially dramatic.

Result-based selection is the one that deserves the strongest scrutiny.

The Full-Series Check

Before accepting a claim about improvement or decline, widen the view.

Find the earliest available measurement.
Find the latest available measurement.
Mark when the claimed cause or intervention began.
Check whether the trend had already started.
Check whether the trend continued, levelled off or reversed.
Check whether the measurement method stayed the same.
Check whether the environment or other important conditions changed.

This does not guarantee a correct causal conclusion. It prevents a short slice from impersonating the whole history.

Worked Case 2: “Our Class Improved by 20 Marks”

A tutoring advertisement says a class improved from an average score of 55 to 75. Ignore the educational marketing question and treat the numbers as a scientific evidence problem.

Suppose the full sequence of average scores across five tests was 48, 52, 55, 68 and 75. The advertised window compares the third and fifth tests.

Several interpretations remain possible:

The class was already improving before the chosen start point.
The later tests may have differed in difficulty.
The group composition may have changed.
The intervention may have helped, but the two numbers alone do not isolate how much.
The final score may be part of a genuine sustained improvement—or a temporary high point. More later evidence would tell us.

The scientific lesson is not “never trust before-and-after numbers”. It is “ask what the boundary hides”.

Temporary Change vs Sustained Change

Time matters because a short-lived response and a stable change can look identical at one moment.

Imagine a plant leaf wilts at noon, recovers in the evening and looks normal the next morning. A photograph taken only at noon could support “the leaf was wilted at noon”. It would not support “the plant permanently deteriorated”.

When the claim is about persistence, the evidence must include enough time to test persistence.

Start Points Can Create Large Percentage Changes

Suppose a value rises from 10 to 20. That is a large relative increase. If the wider record was 30, 25, 10, 20, 22, the dramatic rise from 10 to 20 is still real, but it sits inside a longer pattern that first fell sharply.

The percentage is not wrong merely because the starting point is low. The scientific question is whether that starting point was chosen because it represents the process fairly—or because it creates a dramatic story.

A Window Can Also Hide Delayed Effects

Sometimes the problem is not a dramatic window but a window that ends too early. A treatment may show no immediate effect but a later one. A material may look stable for ten minutes but fail after an hour. A seed may show no visible change on Day 1 but germinate on Day 3.

Therefore, the correct window depends on the process. Scientific caution means neither demanding an infinitely long study nor accepting an arbitrarily short one. It means matching the observation period to the mechanism and question.

What Would Strengthen a Time-Based Claim?

A clearly stated reason for the start and end points.
A full or sufficiently wide time series rather than only two selected points.
Consistent measurement methods across the period.
Information about conditions that changed during the series.
Repeated patterns across comparable trials.
A comparison group or baseline where the causal claim requires one.
Evidence that the effect survives reasonable alternative windows.

What Would Weaken It?

The claim disappears when the graph is extended slightly earlier or later.
The trend had already begun before the supposed cause appeared.
The measuring instrument or procedure changed at the boundary.
The start point is an extreme high or low with no scientific justification.
The period is too short to test the claimed lasting effect.
The report omits later reversal or earlier decline that changes the interpretation.

Do Not Confuse This With the “Axis Starts at Zero” Rule

A graph can exaggerate a visual difference because of its vertical scale, and eduKateSengkang already has a guide on axes that do not start at zero. But this Reality Lab asks a different question. Even with a perfectly labelled axis, a graph can create a selective story by displaying only a narrow time interval.

Scale asks, “How is the size of difference shown?” Time-window selection asks, “Which part of the history was allowed into the graph?”

PSLE-Style Transfer Case

A student measures the temperature of water as it cools. A short graph from Minute 2 to Minute 4 shows a steep fall. The student concludes, “The water always cools at this same rate.”

Why is the conclusion too broad?

Reasoning: the selected window only shows the rate of temperature decrease between Minutes 2 and 4. It does not show that the same rate continues throughout cooling. Additional measurements before and after that interval are needed to determine whether the rate remains the same.

Now transfer the idea back into real life. Any headline that turns “during this period” into “always” has extended the conclusion beyond its time evidence.

Delayed Independent Return

Two days later, find any harmless time graph in a newspaper, school resource, app or public website. Ask:

What is the earliest date shown?
What is the latest?
Why might those boundaries have been chosen?
Would adding earlier or later data change the conclusion?
Does the graph support a temporary statement, a sustained statement or a causal statement?

If you can answer without being told which trick to look for, the skill is becoming scientific judgement rather than a memorised warning.

Useful eduKateSengkang Routes

Parent and Tutor Teaching Guide

When a child sees a dramatic graph, resist starting with “Is this misleading?” That question already suggests the answer. Ask instead, “What period are we looking at?” Then: “What would you want to see before this? What would you want to see after?”

Next ask whether the chosen period has a scientific reason. This is important because not every short window is deceptive. A fair investigation may define a fixed time interval before collecting data. The goal is to help the learner distinguish justified boundaries from result-driven boundaries.

For stronger learners, present the same full data series with three different windows and ask them to write the most defensible conclusion for each. Then reveal the whole series and ask which conclusions still survive.

Authoritative Sources

The Quiet Rule to Keep

A trend lives inside a time boundary. Before believing the story told by the line, inspect the boundary that allowed that story to appear.

The strongest scientific question is often not “Did it change?” but “Would I tell the same story if I could see the rest of the record?”

Small Group Tutorials