How to Read PSLE Science Data With Two Conditions at Once Without Comparing the Wrong Pair

Wait, What? The Biggest Difference in the Table May Be the Wrong Comparison

A results table can show two scientific conditions at the same time. One condition may run down the rows. Another may run across the columns. Every cell then belongs to a particular combination of both conditions.

This creates a dangerous shortcut. Learners often choose two cells with the largest visible difference and compare them. But if both the row condition and the column condition changed, that pair cannot isolate what either condition did.

When two conditions vary, compare cells that keep one condition the same while changing the other.

The table is not just a collection of numbers. It is a map of scientific comparisons.

Quick Answer

When PSLE Science data contains two conditions at once, identify which condition is encoded by the rows and which by the columns. Then decide which condition the question asks about.

Use this route:

NAME ROW CONDITION → NAME COLUMN CONDITION → NAME MEASURED OUTCOME → CHOOSE THE CONDITION TO TEST → HOLD THE OTHER CONDITION FIXED → COMPARE MATCHED CELLS → REPEAT AT ANOTHER MATCHED LEVEL IF NEEDED → DESCRIBE THE PATTERN → EXPLAIN ONLY IF THE SCIENCE AND METHOD SUPPORT IT.

The Exact PSLE Science Learning Job This Guide Owns

This guide owns one learner job: how a Primary 5 or Primary 6 learner reads a two-condition data display and chooses matched comparisons that change one condition while keeping the other one fixed.

It does not own general fair-test theory, generic graph reading or the scientific concepts used in examples. It is narrower: when a table or grouped display deliberately contains combinations of two conditions, how do you avoid comparing a pair in which both conditions changed?

This is also different from the existing guide on several conditions changing at once. That page diagnoses confounded comparisons. This page teaches how to find a clean comparison inside a data grid that already contains multiple combinations.

Why This Matters in the 2026 PSLE Science Frame

SEAB states that the PSLE Science paper for examination from 2026 assesses the 2023 Primary Science syllabus. Candidates are expected to apply scientific knowledge and inquiry, interpret and analyse information, evaluate observations and methods, and communicate explanations and reasoning.

Reading a multi-condition display is an analysis job. You must identify the parts of the information, preserve what each condition means, find a valid comparison and resist a visually dramatic but scientifically unmatched pair.

This guide does not claim that any particular table format must appear in PSLE or that one answer template is officially required.

Build the Cell Address Before Reading the Number

Consider this original practice table:

Condition X	Condition Y = Low	Condition Y = High
X = 1	12	18
X = 2	17	25
X = 3	21	31

The number 25 does not mean anything scientifically on its own. Its full address is:

X = 2, Y = High, measured outcome = 25 units.

Every comparison should preserve these addresses before arithmetic or explanation begins.

To Test the Row Condition, Stay in One Column

Suppose the question asks how X is related to the measured outcome when Y is Low.

Stay in the Low column and compare 12 → 17 → 21 as X changes from 1 → 2 → 3. Y remains Low throughout that comparison.

You may then repeat the same analysis in the High column: 18 → 25 → 31. This lets you ask whether the relationship with X appears under both Y conditions.

To Test the Column Condition, Stay in One Row

Suppose instead the question asks what happens when Y changes from Low to High at X = 2.

Stay in the X = 2 row and compare 17 with 25. X is now held fixed while Y changes.

The same comparison can be repeated at X = 1 and X = 3 if the question asks whether the effect is consistent across different X values.

Why Diagonal Comparisons Are Dangerous

Comparing 12 at X = 1, Y = Low with 31 at X = 3, Y = High changes both X and Y. The outcome difference of 19 units is real, but it cannot be assigned to X alone or Y alone from that comparison.

A diagonal comparison may answer a simple question such as “Which combination gave the larger outcome?” It is not automatically valid evidence for “What is the effect of X?”

The question job decides whether the pair is legitimate.

Worked Example 1 — Temperature and Surface Condition

Original practice situation: a fictional process is tested at three temperatures and with two surface conditions. The measured outcome is recorded after the same time.

If you want to compare temperature, choose one surface condition and compare across temperatures. If you want to compare surface condition, choose one temperature and compare the two surfaces.

Do not compare “low temperature + smooth surface” with “high temperature + rough surface” and then claim temperature caused the whole difference. Both conditions changed.

Worked Example 2 — Light Condition and Water Amount

Imagine a practice table showing plant growth for two light levels and three water amounts. The scientific concept of plant growth remains with its existing owner. Here we focus only on the data logic.

To ask about water amount, compare growth values at different water amounts while holding the light condition fixed. To ask about light, compare the light conditions at the same water amount.

If growth differs across both dimensions, the table can show that the outcome is associated with both tested conditions. Strong causal explanation still requires a suitable investigation design and relevant Science.

Worked Example 3 — A Grouped Bar Display

The same logic applies if the data are shown as grouped bars rather than a table. Suppose each category on the horizontal axis has two bars, P and Q.

First decode what the category means and what P/Q means. Then choose one comparison dimension deliberately:

compare P and Q within the same category to change the series while keeping category fixed;
compare P across categories to change category while keeping the series fixed;
do not compare one P bar in Category 1 with one Q bar in Category 3 and attribute the whole difference to either dimension.

Worked Example 4 — Two Factors Do Not Automatically Mean Interaction

If the difference between Low and High Y is 6 units at X = 1, 8 units at X = 2 and 10 units at X = 3, the effect associated with changing Y is not identical across X levels.

At Primary level, you do not need advanced statistical terminology. You can say carefully that the difference between the Y conditions changes across the tested X values.

Do not jump to a complicated causal model unless the question’s Science supports it. The table shows a pattern; mechanism is another job.

The Four Comparison Types

Comparison	What changes?	What stays fixed?	What it can help answer
Across one row	column condition	row condition	Effect/relationship of column condition at one row level
Down one column	row condition	column condition	Effect/relationship of row condition at one column level
Matched repetition	same chosen condition across several fixed levels	one comparison basis at a time	Whether a relationship is similar across levels
Diagonal	both conditions	neither	Difference between combinations, not isolated effect of one condition

Do Not Turn a Two-Condition Table Into Two Separate One-Condition Experiments

The point of the grid is that each outcome belongs to a combination. If you read row labels without column labels—or vice versa—you lose half the condition.

This is why every value should be read with a full address: row condition + column condition + measured outcome + unit.

Use Several Matched Comparisons Before Claiming a General Pattern

One matched pair tells you what happened at one level of the other condition. If the question asks for a broader relationship, inspect several matched pairs.

For example, if High Y gives a higher outcome than Low Y at X = 1, X = 2 and X = 3, you have stronger evidence that the same direction of difference appears across the tested X levels.

Still keep the conclusion inside the tested conditions. Three rows are not every possible scientific condition.

When One Cell Is Missing

If X = 3, Y = High has no recorded value, do not invent one merely to complete the grid. Use the comparisons that are actually available.

You may be able to compare Y at X = 1 and X = 2 but not X = 3. The missing cell limits the evidence rather than becoming a guessing invitation.

When the Outcome Uses Different Units or Scales

Matched comparisons require the same scientific quantity expressed on a compatible basis. If one part of the display uses a different unit, convert only if the curriculum-level task and information make that appropriate. If different columns measure different outcomes, they may not be direct comparisons at all.

Do not let identical-looking numbers erase different units or quantities.

The Earliest-Weak-Link Diagnostic

Failure signature	Earliest weak link	Repair
Chooses the largest visible difference.	Question variable not identified.	Name the condition being tested before selecting cells.
Compares diagonal cells to explain one factor.	Both conditions changed.	Hold one condition fixed.
Reads a number without its row label.	Cell address incomplete.	Say row + column + outcome aloud.
Uses one pair to claim the pattern always holds.	Evidence coverage too narrow.	Check matched pairs at other levels.
Explains the pattern from the table alone.	Pattern and mechanism collapsed.	Use scientific knowledge and method evidence for cause.
Invents a missing cell.	Grid completeness mistaken for evidence.	Leave unmeasured combinations unknown.

Misconception Repair — “Same Row Means Same Experiment”

A row may contain several experimental combinations because the column condition changes. “Same row” means one encoded condition is held fixed; it does not mean every condition is identical.

Misconception Repair — “The Highest Cell Shows the Best Factor”

The highest outcome belongs to one combination of conditions. It does not by itself tell you which factor caused that outcome or whether that factor is universally best.

Misconception Repair — “Two Factors Mean the Test Is Unfair”

Not necessarily. A well-designed investigation can deliberately test combinations of two conditions. The learner’s job is to choose matched comparisons that isolate the question being asked. What would be weak is using an unmatched pair to claim the effect of only one factor.

The Two-Condition Data Protocol

Read the full title or question.
Name the row condition.
Name the column condition or series.
Name the measured outcome and unit.
Write the full address of one sample cell.
Identify which condition the question wants you to investigate.
Hold the other condition fixed.
Select matched cells.
Compare the correct values.
Repeat at another fixed level when a broader pattern is needed.
State the observed relationship within the tested conditions.
Add a mechanism only when scientific knowledge and method evidence support it.

Original Practice Set

Condition A	B = P	B = Q
A1	8	13
A2	11	18
A3	14	22

To compare P with Q at A2, use 11 and 18.
To inspect the relationship with A under P, use 8, 11 and 14.
Comparing 8 with 22 changes A and B together.
The difference Q − P is 5, 7 and 8 across A1, A2 and A3; it is not constant.

Unfamiliar Transfer Challenge

A fictional device is tested at three settings of X and two settings of Y. You know nothing about the device. The output table is:

X	Y = M	Y = N
1	4	7
2	6	11
3	9	15

Without knowing the topic, you can still make valid structural statements: at every tested X level, N has a higher outcome than M; within either Y setting, the outcome rises across X = 1 to 3. You cannot explain why without a scientific rule or mechanism.

Delayed Independent Return Test

After several days, use a fresh two-condition table or grouped display. Without notes, identify:

row condition;
column condition;
outcome and unit;
one valid row-wise comparison;
one valid column-wise comparison;
one tempting but unmatched diagonal comparison;
one pattern supported by several matched comparisons;
one causal claim the display alone cannot prove.

Answer-Checking Receipt

Did I read both conditions?
Did I keep every value attached to its full address?
Did I identify the condition the question asks about?
Did I keep the other condition fixed?
Did I compare like with like?
Did I check more than one matched pair if claiming a broad pattern?
Did I preserve units?
Did I keep pattern and mechanism separate?
Did I leave missing combinations unknown?

Parent and Tutor Teaching Guide

Draw a 3 × 2 grid and ask the learner to point to two cells that compare Condition X while keeping Y the same. Then ask for two cells that compare Y while keeping X the same.

Next ask the child to choose a diagonal pair and explain why it is weaker for isolating one condition. Do not begin with arithmetic. The comparison structure should become visible before number crunching.

Once this is secure, switch the representation to grouped bars. If the child can preserve the same matched comparisons after the display changes, the reasoning is transferring.

Useful Internal Routes

Authoritative References and Evidence Boundary

The row-and-column method is a comparison scaffold. It does not turn every two-condition data set into proof of two independent causal effects. Method design, measurement quality and the relevant scientific concepts determine what stronger conclusions can be supported.

The Quiet Return

When two conditions share one table, the nearest number is not always your comparison.

Choose the question first. Hold one condition still. Change the other. Then compare.

A crowded grid becomes much simpler when every cell has an address and every comparison has a scientific purpose.

Previous: How to Track What Enters, Leaves and Remains in a PSLE Science System
Next: How to Read a PSLE Science Table With Grouped Headings Without Losing Which Column Means What

Small Group Tutorials