How to Keep a PSLE Science Investigation Record That Preserves What Actually Happened

Wait, What? The Neatest Final Table Can Hide the Most Important Part of the Investigation

A learner carries out a simple investigation. The first reading looks strange, so it is crossed out. Halfway through, the learner notices that one part of the method is difficult and changes it slightly. The final table is copied neatly onto a fresh page. The graph looks smooth. The conclusion sounds sensible.

Then someone asks three ordinary questions:

What was the original method?
Which reading was corrected, and why?
Did the method change before or after the surprising result appeared?

The learner cannot remember.

The problem is not untidy handwriting. The problem is that the evidence history has disappeared.

A useful Science investigation record does more than store the final answer. It preserves enough of what actually happened that the evidence can still be checked later.

That is the job of this guide.

Quick Answer

When you keep a PSLE Science investigation record during learning or practical work, separate the investigation into layers:

Question: What are you trying to find out?
Prediction: What do you expect before seeing the result, and why?
Plan: What will you change, keep comparable, observe or measure?
Method as actually carried out: What did you really do?
Changes: If the method changed, what changed and when?
Raw evidence: What did you actually observe or measure?
Corrections: If a reading was recorded wrongly, preserve the correction trail rather than silently replacing history.
Processed result: What table, comparison, average, graph or derived value did you make from the raw evidence?
Conclusion: What does the evidence support?
Limit: What does the evidence not establish?

The central rule is:

RECORD FIRST → INTERPRET SECOND → REVISE THE EXPLANATION WHEN NEEDED, NOT THE PAST.

The Exact PSLE Science Learning Job This Guide Owns

This guide owns one learner job: how a Primary 5 or Primary 6 learner keeps an investigation record so the question, prediction, actual method, method changes, observations, measurements, corrections, results and conclusion remain traceable to what actually happened.

It is not a rule that PSLE candidates must bring or submit a laboratory notebook. It is not an official answer template. It does not replace the separate guides on designing a results table, writing a reproducible method, handling an anomalous result, correcting a measurement, or keeping a prediction separate from an observation.

Those pages own their specific jobs. This page owns the record across the whole investigation: how the pieces remain connected so another reader—and your future self—can tell what was planned, what happened, what changed, what was observed and what was concluded.

Why This Matters in the Current PSLE Science Frame

For examination from 2026, Standard PSLE Science assesses the 2023 Primary Science syllabus. SEAB’s assessment objectives include applying scientific knowledge and scientific inquiry skills, interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning.

A traceable investigation record is therefore useful training because it makes those inquiry moves visible. It helps the learner distinguish what was given, what was predicted, what was done, what was observed, what was calculated and what was inferred.

This guide does not claim that SEAB requires a particular notebook format, colour code, number of pages, sentence frame or marking template. The record is a learning tool for better scientific reasoning.

The Difference Between a Final Answer and an Evidence Record

A final answer is designed to communicate the conclusion required by a question.

An evidence record has a different purpose. It preserves the path that produced the conclusion.

Final answer	Investigation record
Shows the response you finally give	Shows what happened before the final response
Usually concise	May include raw readings, notes and corrections
Can be rewritten for clarity	Past evidence should not be silently rewritten
Emphasises the conclusion	Preserves provenance: where each claim came from
May omit abandoned ideas	Can preserve relevant method changes or failed checks

This difference matters because a beautiful final answer can be scientifically weak if nobody can tell where its evidence came from.

The Investigation Record Has a Timeline

Science is not only a collection of facts. An investigation unfolds through time.

Before the test, you may have:

a question;
a prediction;
a proposed method;
a recording plan.

During the test, you may have:

actual actions;
time points;
observations;
instrument readings;
unexpected events;
method adjustments;
repeats.

After the test, you may have:

calculations;
tables;
graphs;
comparisons;
a conclusion;
a limitation;
a proposed follow-up.

A good record keeps these stages from collapsing into one another.

Later knowledge should improve your explanation. It should not travel backwards in time and change what you claim you originally observed.

Layer 1 — Write the Scientific Question Before the Evidence Arrives

The question gives the investigation its job.

Suppose you want to investigate how the amount of exposed surface affects the time taken for a wet material to dry under comparable conditions.

Writing the question first helps you later check:

whether the method actually tested that relationship;
whether the recorded outcome answers the question;
whether the conclusion stayed within the tested conditions.

Without the question, a collection of readings can look scientific while answering nothing in particular.

Layer 2 — Preserve the Prediction as a Prediction

A prediction belongs before the result.

Write what you expect and the scientific reason you are using. Then leave that original prediction visible even if the result surprises you.

If the prediction is wrong, that is useful. You can ask whether:

the concept was misunderstood;
the chosen mechanism did not fit the situation;
an important condition was overlooked;
the method did not measure the intended outcome well;
natural or measurement variation affected the evidence.

A rewritten prediction teaches almost nothing because it erases the comparison between expectation and evidence.

Layer 3 — Distinguish the Planned Method From the Method Actually Carried Out

Plans and reality are not always identical.

You may plan to record a temperature every minute but discover that the measuring device takes longer to stabilise. You may plan to use one container and find that it leaks. You may plan to place two set-ups side by side and later notice that one is shaded.

If the method changes, do not quietly rewrite the original plan as though the new method had been used from the beginning.

Record:

what changed;
when it changed;
why it changed;
which results were collected before the change;
which results were collected after the change;
whether the two sets of results are still directly comparable.

This is where record-keeping becomes scientific rather than decorative.

Worked Case 1 — The Measuring Interval Changes Halfway Through

A learner records the temperature of a cooling cup at 0, 2, 4 and 6 minutes. The learner then becomes busy adjusting another part of the apparatus and records at 10 and 14 minutes instead of 8 and 10.

A poor final record might simply list all temperatures in a column and make the spacing look regular.

A better record preserves the actual times. The later four-minute gaps are part of the evidence history.

Why does this matter? Because temperature change over four minutes is not directly the same quantity as temperature change over two minutes. A graph can still plot the actual time points, but the learner must not pretend the sampling interval stayed constant.

Layer 4 — Raw Evidence Comes Before the Neat Summary

Raw evidence is what was directly observed or measured before later summarising.

Examples include:

individual temperature readings;
counts made at each time point;
the exact time an event was first observed;
descriptions such as clear, cloudy, wilted or unchanged when the criterion has been defined;
separate repeat results before an average is calculated.

Keep raw evidence separate from a processed result.

If three trials produce 12 cm, 13 cm and 11 cm, the average is useful. But the average does not replace the three raw values. The spread itself may tell you something about variation.

Worked Case 2 — The Average Hides an Important Difference

Two conditions each produce an average result of 20 units.

Condition P has repeat values of 19, 20 and 21. Condition Q has 10, 20 and 30.

If the learner records only “average = 20” for both conditions, the record loses information about variation.

The correct lesson is not that one condition must be rejected. The lesson is that the processed summary and the raw evidence answer different questions. Preserve both when they matter.

Layer 5 — Keep Time Attached to the Evidence

A reading without a time can become ambiguous in a changing system.

“Temperature = 42°C” means very little if the investigation depends on cooling and the record does not say whether 42°C was measured:

at the start;
after two minutes;
after ten minutes;
after the reading stabilised;
after the set-up was moved.

Time is not always the independent variable. But it can still be essential provenance.

Layer 6 — Keep the Scientific Object Attached to Each Reading

A number must belong to something.

If you have Set-up P, Set-up Q and three specimens in each, record which reading belongs to which object or group. Do not create a list of numbers that can no longer be traced back to their source.

This sounds obvious until labels are reused, samples are moved, containers are swapped or several measurements are made at once.

A practical habit is to keep a simple identity chain:

OBJECT / SET-UP → CONDITION → TIME → MEASUREMENT OR OBSERVATION.

Worked Case 3 — Two Leaves Swap Trays

Two similar leaves, P and Q, are placed under different conditions. During the investigation, the trays are moved and the learner can no longer remember which leaf is P.

The correct response is not to guess from appearance.

The identity chain has been broken. Any later conclusion depending on P versus Q is weakened unless the identity can be recovered from an independent label or record.

A record is useful precisely because it exposes this problem instead of hiding it.

Layer 7 — Corrections Need a Trail, Not a Disappearing Act

People make recording mistakes. A digit may be copied wrongly. A unit may be omitted. A reading may be transposed into the wrong row.

When you know a record is wrong, correct it clearly. But keep enough information to show that a correction occurred.

For learning purposes, a simple approach is:

keep the original entry readable;
mark that it was incorrect;
write the corrected value;
note the reason if it is not obvious;
do not count both old and corrected values as two independent measurements.

This differs from an anomalous result. A surprising measurement is not automatically a recording error. Do not “correct” a result just because it does not fit the pattern you expected.

Worked Case 4 — 73 Becomes 37

A learner writes 73 mL in a table. Immediately after checking the measuring cylinder, the learner notices that the actual reading was 37 mL and that the digits were reversed during copying.

This is a recording correction. The learner can preserve 73 mL as the original transcription, mark it incorrect, and replace it with 37 mL.

Now imagine a different case: the instrument genuinely displayed 73 mL even though nearby repeats are 36, 37 and 38 mL.

That is not automatically a copying error. It is an unusual result that should be investigated, not silently transformed into 37 mL.

Layer 8 — Record Method Changes Before Judging Their Effect

A method change can improve an investigation, but it can also create a comparison problem.

Suppose the first two trials use a ruler with centimetre markings. The last three use a ruler with millimetre markings. The later tool may have finer resolution, but the record should show the change because the evidence was not collected in exactly the same way.

Or suppose the first condition is measured after five minutes while later conditions are measured after ten minutes. The later method might be more convenient, but direct comparison can become invalid.

Do not decide first that the method change “does not matter” and then omit it. Record it first. Evaluate its effect second.

The Difference Between Method Revision and Evidence Revision

It is perfectly reasonable to revise a method for future trials after discovering a problem.

It is not reasonable to revise the historical description of earlier trials so they appear to have used the improved method all along.

Reasonable revision	Misleading revision
“From Trial 4 onward, the probe was held at the same depth.”	Rewriting Trials 1–3 as though probe depth had already been controlled
“The observation interval was changed from 5 min to 2 min after the pilot trial.”	Redrawing the pilot data at 2-min intervals that were never measured
“A leaking container was replaced before the next repeat.”	Keeping the leak secret and presenting all repeats as identical
“The incorrect copied value was corrected after checking the original display/photo.”	Changing a surprising genuine reading because it looks inconvenient

Layer 9 — Distinguish Observation From Interpretation

“The liquid became cloudy” is an observation.

“A new substance must have formed” is an interpretation unless the investigation provides enough evidence for that conclusion.

“The plant grew 3 cm” is a measurement-derived statement.

“The plant grew faster because it received more light” is a causal explanation that requires an appropriate comparison and concept.

A strong record can keep these layers separate without pretending that observations are meaningless. Observations are the evidence surface; interpretations are the reasoning built from them.

Worked Case 5 — “The Seed Is Dead”

A seed has not germinated by the end of the observation period.

The direct observation is that germination was not observed during that period under the tested conditions.

Writing “the seed is dead” into the observation column turns an inference into a recorded fact. The evidence may not justify that conclusion.

A traceable record protects the learner from accidentally making the conclusion stronger than the observation.

Layer 10 — Processed Results Must Point Back to Raw Evidence

Once raw evidence exists, you may process it.

You might calculate:

a change from before to after;
an average of comparable repeats;
a difference between set-ups;
a rate over a stated interval.

You might also create a graph or summary table.

The processed result should remain traceable to the values used. If someone asks “Where did 4.5 come from?”, the record should make it possible to reconstruct the calculation.

Worked Case 6 — A Difference Without Its Starting Values

A learner records only “temperature decreased by 8°C”.

That may be enough for one comparison, but it loses the original starting and final temperatures.

If the later question asks whether two set-ups started at comparable temperatures, the missing raw values matter.

Therefore, when the raw readings are available and relevant, keep them. The derived change is another layer, not a replacement history.

Layer 11 — Conclusions Need an Evidence Address

A conclusion should be able to answer the question:

Which recorded evidence supports this statement?

If the answer is “I just know the Science,” then the statement may be background knowledge rather than a conclusion from this investigation.

If the answer is “because the result in P was 12 and Q was 18 under comparable conditions,” then the record provides an evidence address.

The phrase “evidence address” is an eduKate teaching idea, not an official SEAB term. Its job is simply to force traceability between claim and evidence.

Layer 12 — Record the Limit Beside the Conclusion

A good conclusion says what the evidence supports. A good limit says where that support stops.

If only three temperatures were tested, the record should not quietly transform the conclusion into a universal rule about every possible temperature.

If the result comes from one specimen, do not automatically claim all specimens behave identically.

If the instrument could not detect changes smaller than one unit, do not pretend the data prove there was no smaller change.

Recording the limit is not weakening Science. It is keeping the claim honest.

One Useful Page Layout — Without Turning It Into a Mandatory Template

A learner can divide an investigation page into flexible zones:

Question / purpose
Prediction + reason
Planned variables and method
Actual method notes / changes
Raw evidence
Corrections or unusual events
Processed result
Conclusion
Limit / next check

The exact layout can change. The scientific requirement is not the shape of the page. It is that the record preserves the important distinctions.

Failure Signature 1 — The Final Table Is Too Perfect

Perfectly smooth data are not automatically suspicious, but a record that contains only the final clean table can hide whether:

raw readings were removed;
repeats differed;
anomalies were discarded;
units changed;
method changes occurred.

Repair: keep the raw observations or measurements from which the neat table was built.

Failure Signature 2 — The Prediction Magically Matches the Result

If every prediction is rewritten after data collection, the learner loses a powerful diagnostic tool.

Repair: date or position the prediction before the result and leave it unchanged. Add a later note explaining what the evidence taught you.

Failure Signature 3 — “We Changed Something, But I Forgot What”

This often appears when a method becomes difficult in real use.

Repair: record the adjustment at the moment it happens, even briefly. A short note such as “Trial 3 onward: container replaced because original leaked” is more useful than a perfect memory invented later.

Failure Signature 4 — The Record Contains Conclusions in the Observation Column

Examples:

“lost heat faster” instead of recording the relevant temperature change;
“less photosynthesis” instead of recording the observed indicator used in that investigation;
“stronger magnet” instead of the measured or observed outcome used to compare magnetic effect.

Repair: write what was directly observed or measured first. Put the scientific interpretation later.

Failure Signature 5 — A Correction Becomes an Extra Trial

If a copied value of 73 is corrected to 37, those are not two independent readings.

Repair: preserve the correction history but count only the valid measurement in the evidence summary.

Failure Signature 6 — The Learner Cannot Reconstruct the Graph

If the graph exists but the table or raw observations used to make it are gone, checking becomes harder.

Repair: keep the source values and units attached to the graph. A graph is a representation of evidence, not a replacement for the record that produced it.

The Earliest-Weak-Link Diagnostic

What you see	Earliest weak link	Repair
Prediction altered after the result	Expectation and evidence not separated	Freeze prediction before data collection
Method change missing	Planned method treated as actual method	Record what changed and when
Only averages remain	Raw evidence discarded	Keep individual comparable repeats
Observation includes a causal explanation	Evidence and inference mixed	Record observable/measurable result first
Corrected value counted twice	Evidence history confused with evidence quantity	Mark replacement clearly
Graph cannot be traced to source data	Processed result detached from provenance	Keep table/raw values and units
Conclusion has no supporting line of evidence	Claim–evidence link missing	Give conclusion an evidence address
All tested cases become a universal rule	Conclusion scope exceeds evidence	Record tested range and limit

Misconception Repair — “A Science Notebook Is Just Notes About Science”

A topic notebook can contain definitions, diagrams and explanations. An investigation record has another job: preserve the provenance of evidence.

You can use the same book for both, but label the functions clearly. A copied definition is not an observation. A model answer is not a measurement. A prediction is not a result.

Misconception Repair — “If the Final Answer Is Correct, the Record Does Not Matter”

A correct conclusion reached for the wrong reason can hide a weak understanding. A traceable record allows you to inspect the reasoning path.

This matters especially in practice. You want to know whether success came from:

a sound prediction;
a suitable method;
careful observation;
correct data handling;
good scientific explanation;
or a lucky guess.

Misconception Repair — “Never Change Anything in a Record”

That is too rigid.

You may need to correct a transcription error, add a missing unit, clarify a label or improve the method for a later trial.

The key is traceability: make the correction visible and do not disguise a later improvement as an earlier fact.

Misconception Repair — “Every Small Detail Must Be Recorded”

A record can become unusable if it contains endless irrelevant detail.

Record what helps establish:

what was tested;
what conditions mattered;
what was actually done;
what was observed or measured;
what changed;
where a result came from;
what conclusion the evidence can support.

The goal is not maximum writing. It is sufficient scientific traceability.

Evidence Versus Memory

Human memory is useful but reconstructive. After we know the result, it becomes easy to remember our earlier expectation as more accurate than it really was, or to forget small method changes that later seem unimportant.

An investigation record reduces this problem by moving important information out of memory and onto the page while it is still fresh.

This is one reason the record should be made during the process rather than reconstructed entirely at the end.

Evidence Versus Inference

Use two questions:

What did I actually observe or measure?
What do I think that evidence means?

If those two answers are identical every time, inspect your record. You may be writing interpretations as though they were observations.

Evidence Versus Model

A scientific model can help explain the evidence, but the model is not the evidence itself.

For example, a learner may use a particle model, a force diagram or an energy-flow idea to explain a pattern. Keep the model connected to the observations it is trying to explain. If a different model could also fit, the record should not pretend the explanation was directly seen.

How a Traceable Record Helps When the Result Is Unexpected

Unexpected results are where good records become especially valuable.

You can check:

Was the prediction based on the correct mechanism?
Were the set-ups actually comparable?
Did the intended condition reach the specimen?
Was the same measurement method used throughout?
Was there a recording or copying error?
Did an apparatus problem occur?
Was the unexpected value repeated?
Is the result genuinely possible under the tested conditions?

Without a record, these become guesses about the past.

How a Traceable Record Helps With Fair Testing

A learner may intend to keep conditions comparable but accidentally change one.

If the record shows starting temperature, specimen identity, duration, measurement position or another relevant condition, you can see whether the fair comparison survived in practice.

This is stronger than writing a generic sentence such as “keep all variables the same.” A record can show whether a relevant condition actually stayed comparable.

How a Traceable Record Helps With Repeated Trials

Repeated trials only help if the record lets you tell which results belong to which repeat and whether the method was comparable.

Label Trial 1, Trial 2 and Trial 3, or another clear identifier. Keep individual readings until you have decided that averaging or another summary is justified.

Do not let “repeat three times” become a ritual. The scientific job is to gain additional evidence under comparable conditions.

How a Traceable Record Helps With a Follow-Up Investigation

A follow-up investigation should be able to say what it is responding to.

Maybe the first investigation showed a surprising value. Maybe two explanations still fit. Maybe the tested range was too narrow. Maybe a measurement method could not resolve a small difference.

A good original record lets the follow-up target the actual uncertainty instead of starting from a vague memory that “something went wrong.”

Worked Case 7 — One Investigation, Two Later Stories

Two learners run the same drying investigation.

Learner A writes only the final averages.

Learner B keeps the original question, times, individual measurements, a note that one sample was accidentally moved after 12 minutes, and the final averages.

Both learners can produce a graph. But only Learner B can later ask whether the moved sample should be interpreted differently, whether the pattern remains without it, and whether a repeat would resolve the uncertainty.

The difference is not that Learner B wrote more words. The difference is that Learner B preserved the evidence history.

Worked Case 8 — The Method Improves After a Pilot

A learner first tests a method and discovers that five-minute intervals are too wide to see when a rapid change occurs. The learner then uses one-minute intervals in the main investigation.

Do not mix the pilot readings into the final evidence as though they came from the same recording design.

The pilot can remain valuable because it explains why the final method uses one-minute intervals. The record should identify it as preliminary rather than final evidence.

Worked Case 9 — A Photograph Supports, But Does Not Replace, the Record

A learner photographs a set-up at the end of an investigation.

The photograph may preserve useful visual evidence, but it may not show:

the starting condition;
the timing;
the scale;
what changed during the method;
which measurement was made where;
what happened between photographs.

Treat the photograph as one evidence source whose meaning depends on labels, timing and context.

Worked Case 10 — A Smooth Graph From Messy Repeats

A learner averages several repeats and obtains a smooth increasing graph.

The graph can be a useful summary. But if one condition had wildly varying repeats, that variation may matter when judging evidence strength.

Keep the repeat values available. A graph can simplify the evidence without erasing its history.

The Reconstruction Test

Close the investigation for a day. Return later and try to reconstruct it from the record alone.

Can you answer:

What was the question?
What did I predict before the result?
What was changed deliberately?
What was measured or observed?
What important conditions were kept comparable?
What procedure was actually used?
Did the method change?
Which values are raw readings?
Which values are calculations or averages?
Which entries were corrected?
Which results belong to which trial or specimen?
What evidence supports the conclusion?
Where does the conclusion stop?

If several answers depend on memory rather than the page, the record has a gap.

The Provenance Test

Choose any important number, graph point or conclusion and ask:

Where did this come from?

You should be able to trace:

CLAIM → PROCESSED RESULT → RAW EVIDENCE → TRIAL / OBJECT / TIME / METHOD.

If the chain breaks, the result may still be correct, but checking it becomes harder.

The Hindsight Test

Ask:

Did I rewrite my prediction after seeing the result?
Did I remove an inconvenient reading without a recorded reason?
Did I describe the improved method as though it had been used from the start?
Did I change an observation into an interpretation after learning the answer?

If yes, repair the record so the timeline becomes visible again.

The Minimum Sufficient Record

You do not need a novel about every experiment.

A minimum sufficient record usually lets another reader identify:

the question;
the relevant set-ups or specimens;
the condition changed or compared;
the outcome observed or measured;
important timing and units;
the actual method;
raw evidence;
important deviations or corrections;
the processed result;
the conclusion and its limit.

Sometimes fewer fields are enough. Sometimes more context is required. The scientific test is whether the evidence can still be interpreted correctly.

Practice Task 1 — Rebuild a Missing History

Create an original final graph with five plotted points but no raw table. Give it to a learner and ask what information would be needed to audit the graph.

A strong response may ask for:

axis quantities and units;
the measured values;
the tested conditions;
whether points are individual readings or averages;
how many repeats produced each point;
whether the same method was used for all points.

Practice Task 2 — Separate Four Kinds of Statement

For each original scenario, label each sentence as one of:

prediction;
observation / measurement;
processed result;
interpretation / conclusion.

Example:

“I expect P to cool more slowly because…” → prediction.
“P was 64°C after 5 minutes.” → measurement.
“P decreased by 16°C.” → processed result if calculated from readings.
“Under these tested conditions, P lost less thermal energy to the surroundings over the interval than Q.” → interpretation that must fit the supplied evidence and concept.

Practice Task 3 — Find the Moment Comparability Breaks

Write a short investigation record in which one relevant condition changes during Trial 4. Ask the learner to mark the exact point after which the earlier and later data may no longer be directly comparable.

This trains the habit of treating method history as part of evidence quality.

Practice Task 4 — Correct Without Erasing

Give the learner a table containing one obvious transcription error and one merely surprising but possible result.

Ask the learner to decide which can be corrected from independent evidence and which must remain as an anomalous observation until investigated.

Practice Task 5 — Give Every Conclusion an Evidence Address

For five conclusions, ask the learner to point to the exact observation, table row, graph region or comparison supporting each one.

If a conclusion has no evidence address, ask whether it is background scientific knowledge, an unsupported guess, or an overextended claim.

Unfamiliar Transfer Challenge

A learner investigates how one condition affects the time taken for an event to occur.

The record shows:

Condition A: 42 s, 44 s, 43 s;
Condition B: 31 s, 30 s;
a note beside the third planned B repeat: “timer button failed; trial restarted with a different timer”;
Condition B later: 29 s, recorded with the new timer;
a final table showing only A average = 43 s and B average = 30 s.

Ask:

Which values are raw evidence?
Is the failed timing attempt a valid B result?
Does the change of timer affect comparability?
Should the new 29 s reading automatically be combined with the earlier B readings?
What additional information would help you decide?
What conclusion can safely be made from the record as written?

There is no need to force one universal answer to question 3 or 4. The point is to identify the evidence problem: a measurement method changed, so the learner should check whether the two timers are suitably comparable before pooling the readings.

Delayed Independent Return

Three to five days later, give the learner an unfamiliar investigation record and no checklist.

Ask the learner to identify:

the scientific question;
prediction;
actual method;
raw evidence;
processed results;
corrections;
method changes;
claim–evidence link;
conclusion limit.

Then ask one final question:

What part of this investigation could you not reconstruct confidently from the record?

That answer reveals the next recording skill to repair.

The Answer-Checking Receipt

Did I state the investigation question clearly?
Did I keep the original prediction separate from the result?
Did I distinguish the planned method from what actually happened?
Did I record important method changes and their timing?
Did I preserve raw observations or measurements?
Did I keep object, set-up, condition, unit and time attached to readings?
Did I distinguish a correction from an extra repeat?
Did I keep an anomalous result unless there is evidence it was recorded or measured wrongly?
Can I trace processed values back to raw evidence?
Can I trace the conclusion back to the relevant evidence?
Did I keep observation separate from later interpretation?
Did I state the limit of the conclusion?
Could I reconstruct the investigation after a delay without relying mainly on memory?

Parent and Tutor Teaching Guide

Do not begin by giving the child a complicated notebook template.

Begin with one small investigation or original data scenario and ask the child to keep only four things separate:

what I expected;
what I did;
what I observed;
what I think it means.

Once those four are stable, add raw measurements, corrections, repeats, method changes and conclusion limits.

A useful teaching move is to introduce one deliberate interruption. For example, change the measuring tool after two trials or reveal that one sample label fell off. Ask the learner what the record must preserve before deciding whether the evidence is still usable.

Another useful move is to compare two records of the same investigation:

Record A is beautiful but contains only the final averages.
Record B is concise but preserves raw results, timing and one method change.

Ask which record is easier to audit and why.

When the child makes a mistake, avoid turning the page into punishment. The aim is to build a scientific habit: preserve the evidence, locate the break, repair the method or explanation, and test again.

What This Guide Does Not Own

This page does not become the canonical owner for every skill mentioned inside it.

How to design a results table belongs to the results-table guide.
How to write a reproducible method belongs to the method-writing guide.
How to distinguish prediction from observation belongs to the prediction/recording-integrity guide.
How to handle anomalous data belongs to the anomaly guide.
How to correct a measurement without counting the old value twice belongs to the corrected-measurement guide.
How to decide when to average repeats belongs to the averaging guide.
How to evaluate a method flaw belongs to the method-evaluation guides.

This page connects those jobs through one higher-level learner question: Can the investigation still tell its own history accurately enough to be checked?

Useful Internal Routes

Authoritative Sources

These official sources establish the current syllabus and assessment frame. The investigation-record method in this guide is an eduKate learning approach for practising scientific inquiry and evidence discipline; it is not presented as an official SEAB notebook format.

The Quiet Ending

Science is not improved by pretending the path was cleaner than it was.

A wrong prediction can teach you.

A method change can teach you.

An awkward reading can teach you.

A correction can teach you.

But only if the evidence history survives long enough to be examined.

Keep the question. Keep the timeline. Keep the raw evidence. Mark the changes. Trace the conclusion back to what actually happened.

That is not extra paperwork around Science.

That is one of the ways a learner begins to think scientifically.

Small Group Tutorials