Series ID: PSLE-SCI-REALITY-0448
A scientific chart shows three coloured bands stacked on top of one another across time. The green band seems to bend downward. At the same time, the lower blue band is rising. A learner points at the green band’s upper edge and says, “Green decreased.” That conclusion may be wrong even though the line really does slope downward.
The reason is simple once you reopen the chart: in a stacked area chart, most components do not begin at zero. They sit on top of whatever is below them. The scientific quantity represented by a middle band is usually the vertical thickness of the band—the difference between its upper and lower boundaries—not the height of its upper boundary above zero.
This makes stacked area charts a useful PSLE Science Reality Lab object. They compress a total and its parts into one picture, but the picture can tempt the eye to track the wrong line. The learner’s job is to recover each component from the representation before deciding whether it increased, decreased or stayed the same.
Wait, What? The Top Edge Can Fall While the Band Itself Grows
Imagine an original chart showing the daily mass of three materials collected by a recycling study. Blue is paper. Green is plastic. Orange is metal. The chart stacks the materials so that the total height shows total mass collected.
On Day 1, paper is 20 kg and plastic is 10 kg. The lower edge of the green band therefore begins at 20 kg and the upper edge is at 30 kg. On Day 2, paper falls to 10 kg while plastic rises to 15 kg. The green band now begins at 10 kg and ends at 25 kg.
Look only at the green band’s upper edge: it fell from 30 to 25. You might say plastic decreased. But plastic actually rose from 10 to 15 kg. Its upper edge moved downward because the paper underneath it fell even more strongly.
That is the Reality Lab trap: a visible boundary can move because the component itself changed, because components below it changed, or because both changed.
Quick Answer
Not necessarily. For a middle component in a stacked area chart, do not read its value from the height of one boundary above zero. Recover the component from the difference between its upper and lower boundaries at the same time point. If exact values or a source table are available, use them. The total stack can rise while one component falls, a boundary can fall while a component rises, and a band can appear visually compressed because other components change around it.
The Owned Learner Job — and the Boundary
This article owns one narrow transfer job: evaluate an individual component inside a stacked scientific area chart by separating the component’s own thickness from the cumulative height of the stack. It does not re-own graph scales, axes, percentages, totals, rates or trend interpretation in general.
- For scales, units and reading plotted values, use How to Read Units, Scales and Measurement Resolution Before Using PSLE Science Data.
- For building faithful graphs from tables, use How to Turn a PSLE Science Results Table Into an Honest Graph Without Distorting the Data.
- For a different representation trap involving two scales, use Reality Lab Vol No.030.
Stacked area charts are not “bad charts.” They are especially useful when the reader needs to see how a total changes while also seeing its composition. The difficulty appears when we ask a floating middle band to behave like an ordinary line graph.
What a Stacked Area Chart Is Actually Doing
Suppose three components are A, B and C. At one time point their values are 20, 10 and 5 units.
| Boundary | Height above zero | What it means |
|---|---|---|
| Top of A | 20 | A |
| Top of B | 30 | A + B |
| Top of C | 35 | A + B + C, the total |
Only the bottom component has a zero baseline. The top of B does not mean B = 30. It means A + B = 30. To recover B, subtract the bottom of B from the top of B: 30 − 20 = 10.
The top boundary of the whole stack is special too. It directly shows the total of all components. That is why stacked area charts often make the overall trend easy to see while making individual middle-series trends harder to compare.
Observed, Represented, Claimed and Inferred
| Layer | Example |
|---|---|
| Observed or estimated data | Component values recorded for each year |
| Represented | Values accumulated vertically into coloured bands |
| Claimed | “The total rose over the decade” |
| Inferred | “The green component fell because its upper boundary slopes down” |
The total claim may be easy to read from the top of the stack. The component claim needs a different operation: inspect both boundaries or return to the source values.
Case File A: The Boundary Falls but the Component Rises
Use this original two-day dataset:
| Day | Blue component | Green component | Orange component | Total |
|---|---|---|---|---|
| 1 | 20 | 10 | 5 | 35 |
| 2 | 10 | 15 | 5 | 30 |
The green band runs from 20 to 30 on Day 1 and from 10 to 25 on Day 2. Its upper boundary falls from 30 to 25. Yet its thickness grows from 10 to 15. Green increased.
This example matters because it proves that the slope of a floating boundary is not the same thing as the trend of the middle component. The boundary is cumulative: it carries changes from everything underneath.
Case File B: The Upper Boundary Rises but the Component Falls
Now reverse the trap. On Day 1, Blue = 10 and Green = 20, so the upper edge of Green is at 30. On Day 2, Blue rises to 25 while Green falls to 15, so the upper edge of Green is at 40.
The upper boundary rises from 30 to 40, but Green falls from 20 to 15. The chart has not contradicted itself. Blue underneath Green increased strongly enough to lift the cumulative boundary.
Case File C: The Band Looks Thinner Only Because the Chart Is Small
A science dashboard is viewed on a phone. Four stacked components squeeze into a small vertical space. The green band appears to thin from one month to the next, but the source values are 42 and 41 units.
The one-unit decrease is real, but the visual impression may exaggerate it. Exact values, labels or the underlying table are more reliable for close comparisons. A chart that is good for showing overall composition may be poor for estimating a small change in a middle band.
Case File D: A 100% Stacked Area Chart Changes the Question
Some stacked area charts force the whole stack to 100% at every time point. In that chart, band thickness represents each component’s share of the total, not its absolute amount.
Imagine total observations increase from 100 to 200. Green rises from 40 observations to 60. Its absolute amount increases by 20. But its share falls from 40% to 30%. In a 100% stacked chart, the green band gets thinner even though the number of green observations increased.
This is a second evidence-transfer job inside the same communication object: always ask whether the chart shows absolute values or proportions. The percentage sign is not decoration; it changes the scientific question being answered.
The Representation Check
- Is the chart stacked or are the series merely overlapping?
- Does the vertical axis show original units or percentages?
- Does the whole stack represent a meaningful total?
- Which component sits on the zero baseline?
- For the component of interest, where are its lower and upper boundaries?
- Can exact values be read from labels or a table?
- Has the order of the stacked components changed between panels?
- Are missing values or interpolated periods marked?
The UK Office for National Statistics warns that components in stacked charts can be difficult to compare when they do not share a common baseline. CDC’s stacked-area guidance similarly notes that these charts can communicate the whole well while misleading readers about the progress of individual groups. Those are design observations. The learner’s scientific response is to recover the actual component values before making a trend claim.
The Baseline Check: Which Line Is Zero for This Component?
In an ordinary line graph, every series is usually read against the same numerical axis. In a stacked area chart, the bottom component still starts from zero, but the next component begins at the top of the first, the third begins at the top of the first two, and so on.
That means a floating band has a moving baseline. To know Green at time t, you need:
Green value = height of Green’s upper boundary − height of Green’s lower boundary.
This is not a special PSLE formula. It is simply unpacking what stacking means. If a source table is provided, reading the table is often easier and more accurate.
The Comparison Check: Total, Component or Share?
| Question | Where to look |
|---|---|
| Did the total increase? | Top boundary of the full stack |
| Did the bottom component increase? | Its top boundary, because its bottom is zero |
| Did a middle component increase? | Difference between its upper and lower boundaries |
| Did the component’s share increase? | Use a 100% stack or calculate component ÷ total |
| Which exact value is larger? | Prefer labels or the underlying data table |
The Method and Variable Check
A clear stacked chart can still support a weak scientific claim if the underlying measurements are not comparable. Ask whether each series uses the same unit, time interval and measurement method. If one category is counted weekly and another monthly, stacking them as though they were comparable parts of one total can be scientifically misleading.
Also check whether categories are mutually exclusive. If one observation can belong to two categories at once, adding the components can double-count evidence. A stacked chart visually invites the reader to treat the components as parts of one total, so the data structure must justify that addition.
Alternative Explanations for a Changing Band
- The component itself changed.
- A component below it changed and moved both boundaries.
- The total changed while the component’s share stayed similar.
- The chart is normalised to 100%, so only proportions are visible.
- A category definition changed between periods.
- Missing observations altered the total.
- The stack order changed, making visual tracking harder.
- The chart was resized, making small thickness differences look stronger or weaker.
Again, these are possible explanations, not ready-made conclusions. Good scientific reasoning identifies what additional evidence would separate them.
What Evidence Would Strengthen a Component-Trend Claim?
- The underlying data table with exact component values.
- Direct labels for important values.
- Consistent category definitions and units across time.
- A clearly identified absolute or 100% scale.
- Consistent stacking order.
- A companion line chart or small-multiple chart when individual trends are important.
- Transparent notes about missing, estimated or revised values.
What Evidence Would Weaken It?
A claim such as “Green fell sharply” is weakened when it is based only on a floating upper boundary, when the source table shows little or no decrease, when categories change meaning, when the chart switches between absolute and percentage stacking, or when a large movement in lower components explains the boundary shift.
None of this means the chart must be rejected. It means the conclusion should be tied to the component value rather than to the most eye-catching edge.
How Far Can the Conclusion Travel?
| Claim | Evidence needed |
|---|---|
| The total increased. | Read the top of the full stack, with comparable units and periods. |
| The green component increased. | Recover green’s thickness or use exact green values. |
| Green became a smaller share of the total. | Compare proportions, not just absolute thickness on a changing total. |
| Green caused the total to increase. | Needs causal evidence beyond the chart. |
| Green will continue rising. | Needs evidence about future conditions or a justified model. |
| The chart proves a measurement method is accurate. | No. Representation does not establish measurement accuracy. |
Tempting but Invalid Reasoning
- “The upper edge slopes down, so the component decreased.” Subtract the lower boundary first.
- “The band got thinner, so the absolute amount decreased.” In a 100% stack, a thinner band can mean a smaller share even while the absolute amount rises.
- “The top band is highest, so it has the largest value.” Its position includes every component below it.
- “The chart is stacked, so every category must be part of one valid total.” Check whether categories are mutually exclusive and measured on the same basis.
- “If the total rises, every component rose.” Some components can fall while others increase by more.
- “A colourful smooth shape is evidence of a smooth process.” The drawing connects discrete observations; inspect the measurement frequency and raw data.
Worked Case: Energy Sources in a Fictional School
A fictional school tracks daily energy supplied by three sources. This example is constructed for learning and does not describe a real school.
| Day | Solar / units | Grid / units | Battery / units | Total |
|---|---|---|---|---|
| Mon | 20 | 50 | 10 | 80 |
| Tue | 30 | 45 | 10 | 85 |
| Wed | 40 | 35 | 15 | 90 |
| Thu | 50 | 25 | 20 | 95 |
Suppose the stack order is Solar at the bottom, Grid in the middle and Battery on top. The upper boundary of Grid equals Solar + Grid: 70, 75, 75, 75. A learner looking only at that boundary might say Grid stayed almost constant.
But Grid actually fell from 50 to 25. Solar rose at the same time, keeping the cumulative Solar + Grid boundary nearly flat. This is exactly the kind of hidden movement a stacked representation can create.
The total still rises from 80 to 95 because Solar and Battery increases more than offset the Grid decrease. One chart therefore contains three different stories: the total rises, Solar rises, Grid falls. Reading only one boundary can collapse those stories into the wrong one.
PSLE-Style Transfer Case: The Water-Tank Study
This is an original practice case, not an examination question. Three sources add water to a tank during four equal time periods: rain collection, recycled water and mains water. A stacked area chart shows the amount supplied during each period.
| Period | Rain / L | Recycled / L | Mains / L |
|---|---|---|---|
| 1 | 10 | 20 | 30 |
| 2 | 20 | 20 | 25 |
| 3 | 30 | 20 | 20 |
| 4 | 40 | 20 | 15 |
Rain is at the bottom, recycled water is in the middle. The upper boundary of recycled water rises from 30 to 60 because Rain below it rises. Does recycled water increase?
Answer: No. Its thickness remains 20 L in every period. The rising upper boundary belongs to Rain + Recycled together.
Second question: The top of the whole stack rises from 60 to 75. Does that mean mains water increased?
Answer: No. Mains water actually decreases from 30 to 15 L. The total rises because rain collection increases by more than mains water decreases.
Third question: What conclusion is secure?
The total supplied per period increased from 60 to 75 L; rain contribution increased; recycled contribution stayed constant; mains contribution decreased. Those statements come from the component values, not from guessing which coloured edge points upward.
Delayed Independent Return
Now imagine a different chart. The purple band sits between yellow below and grey above. Purple’s lower boundary rises from 40 to 70. Its upper boundary rises from 55 to 82. Did Purple increase?
Recover its thickness: 55 − 40 = 15 at the first point; 82 − 70 = 12 at the second. Purple decreased from 15 to 12 even though both of its visible boundaries rose.
Explained Practice
1. A middle band runs from 30 to 50 at Time A and from 45 to 65 at Time B. Did it change?
Answer: No. Its thickness is 20 at both times, even though both boundaries moved upward.
2. A band’s upper edge is flat, but its lower edge rises. What happened to the component?
Answer: Its thickness decreased. If the upper boundary stays fixed while the lower boundary rises, the distance between them becomes smaller.
3. The full stack rises from 100 to 120. Can every component have decreased?
Answer: No, not if the components are the complete non-negative parts of that total and are measured comparably. At least enough component value must have increased to raise the total. But some individual components can still decrease.
4. In a 100% stacked chart, Green falls from 40% to 30%. Must its absolute amount have fallen?
Answer: No. The total may have increased. Green can become a smaller share of a larger total while its absolute amount increases.
5. Which component is usually easiest to compare across time in a standard stacked area chart?
Answer: The bottom component because it shares a fixed zero baseline. The total is also easy to read from the top of the entire stack. Middle components have moving baselines.
6. The chart is hard to read but the source table is available. Which should you use for an exact comparison?
Answer: Use the table for exact values and the chart for broader pattern and composition. Choosing the representation that best answers the question is part of evidence literacy.
Model and Measurement Limits
A stacked area chart cannot add certainty that is absent from the measurements. If the underlying values are estimates with uncertainty, the coloured boundaries remain estimates. If sampling is incomplete, a smooth area does not turn the missing periods into observations. If one category was redefined, the continuous band can visually hide a break in comparability.
Some official data-visualisation guidance recommends alternative forms when individual components are the main story because a common baseline makes comparisons easier. That is not a rule that stacked area charts are forbidden. It is a reminder to match the display to the question.
Parent and Tutor Teaching Guide: The Two-Ruler Repair
Draw three stacked bands on graph paper. Give the learner two rulers or two fingers. Place one on the upper edge of the middle band and one on the lower edge. Ask, “Which distance belongs to Green?” This physical move prevents the child from treating the upper edge as a zero-based line.
Next, give two time points where both Green boundaries rise but Green thickness falls. Ask the learner to calculate both differences. Then give the reverse case where the upper edge falls but thickness rises. These two counterexamples are more powerful than a long warning because they break the false rule from both directions.
After that, switch to a 100% stacked chart. Ask, “Is this showing amount or share?” Require the learner to say the denominator. If the total is not visible, provide a second table so the learner can discover that a smaller share can still contain a larger absolute quantity.
Finally, remove the colours and use neutral labels A, B and C. Transfer is complete when the learner no longer depends on familiar topic cues and automatically asks for the component’s own thickness or exact value.
Authoritative Sources and Further Reading
- Singapore Examinations and Assessment Board — 2026 PSLE Science syllabus and assessment objectives.
- Ministry of Education Singapore — 2023 Primary Science syllabus.
- US Centers for Disease Control and Prevention — COVE stacked area chart guidance. CDC notes that stacked area charts communicate the whole across time but can mislead when readers try to follow individual groups.
- UK Office for National Statistics — Choosing a chart type. ONS explains that individual components in stacked charts are harder to compare when they lack a common baseline.
- UK Office for National Statistics — Stacked bar charts, for the closely related principle that stacked segments show parts of a whole and middle segments are difficult to compare precisely.
- UK Office for National Statistics — Axes and gridlines, for accurate scale and baseline communication.
These are communication and evidence sources, not PSLE marking schemes. The examination transfer is the scientific habit: interpret the representation carefully, evaluate what it can support and explain the conclusion using the actual evidence.
Quiet Return: Measure the Band, Not the Drama of Its Edge
A stacked area chart is built by placing one quantity on top of another. That design is excellent for showing the whole. It also means that a middle band inherits the movements of the layers below it.
So when the green edge rises, falls or bends sharply, do not decide too quickly what Green did. Find its lower boundary. Find its upper boundary. Recover the thickness—or use the original table. Then check whether you are reading amount or share.
The quiet scientific habit is this: when a representation stacks evidence, unstack the evidence before you explain it.
